XCAT_Release_Notes

2.9.1 Release Notes


New Function and Changes in Behavior

  • RHEL 7.1 support on x86_64, ppc64 and ppc64le platforms
    • Diskful and diskless
    • Running xCAT management node on RHEL 7.1
  • Ubuntu 14.04.02 support on x86_64 and ppc64le platforms
    • Diskful and diskless
    • Running xCAT management node on Ubuntu 14.04.02
    • Nvidia GPU configuration on IBM Power 8 servers is not support
  • Support Local Mirror for Ubuntu diskfull and diskless OS deployment. This is useful in the case your xCAT MN cannot access internet or the network bandwidth is not good.
  • Ubuntu hierarchy
    • MariaDB is the only supported database for Ubuntu hierarchy
  • SLES 12 diskless support
  • Energy management for IBM Power 8 servers
    • Power consumption information
    • Hardware vitals: temperature, fanspeed, CPU speed, etc.
    • Power saving
    • The energy management for IBM Power 8 servers uses a new mechanism, it does not depends on xCAT Energy Management Plug-in xCAT-pEnergy any more.
    • The setting operation for IBM Power 8 server is only supported for the server which is running in PowerVM mode, Do NOT run the setting for the server which is running in OPAL mode.
  • RHEL 7.1 LE -> RHEL 7.1 BE mixed cluster support
    • Manage and deploy RHEL 7.1 BE compute nodes from RHEL 7.1 LE management node
    • RHEL 7.1 BE -> RHEL 7.1 LE mixed configuration should also work, but was not formally tested
  • Node attribute primarynic is deprecated, installnic could be used to specify the installation nic, if it is blank, the installnic will be default to the nic associates with the mac address specified by the mac attribute.
  • New column nicextraparams is added to the nics table, which could be used to specify arbitrary nic configuration parameters work with confignics postscript.
  • Docker support in xCAT(experimental)
  • Using confluent to replace conserver(experimental). A document was added to describe how to setup xCAT-confluent as xCAT console server.
  • To avoid the Poodle Attack, set TLSv1 as default SSL version for all the SSL connection in xCAT.
  • The GPG key for xCAT pkgs has been changed to 'xCAT Security Key'.
  • Enhance the command restartxcatd to support the 'fast restart xcatd' for systemd enabled system.
  • Support two new netboot methods: grub2-tftp and grub2-http. They can be used to control the communication protocol for grub2.
  • Add the rh7.0 directory in the xCAT dependency tar file. To install xCAT, if the OS of xCAT management node is rh7.0, use the rh7.0 directory in dependency tar file. And use the rh7 directory if the xCAT MN is rh7.1 and higher.

Test Environment

Key Bug fixes

Get all the bugs which were fixed in 2.9.1 release from 2.9.1 defects.

Restrictions and Known Problems

  • "syncfiles" postscript does not work for file syncing during rh7 installation, but it works fine in updatenode process. The workaround is to prefix the "/mnt/sysimage" to the destination directory. See defect 4579 for details.
  • For some P8LE bare metal machine, when using "rsetboot <node> net" to start network install, the installing will hang at petitboot screen after installation. The workaround is to reboot the node with command "rsetboot <node> hd" and "rpower <node> reset". See defect 4611 for details.
  • Postscript 'confignics -s' does not work for sles12. See defect 4565 for details.
  • For Redhat7.0 provisioning, there might be some warnings on deprecated kernel/anaconda options. See defect 4613 for details.
  • Sles12 xCAT management node does not enable the UDP port to receive syslog request from compute node, so the syslog on compute node can NOT be pushed to xCAT MN. The workaournd is to enable the following two lines in /etc/rsyslog.d/remote.conf on xCAT sles12 MN. See defect 4478.
    $ModLoad imudp.so
    $UDPServerRun 514
  • If the /install on the management node is a nfs mounted directory, genimage for the RHEL 7.x diskless image may fail with error "cpio: cap_set_file failed - Operation not supported", see bug 4654 for more details. The workaround is to change the osimage attribute rootimgdir to local file system, run genimage, and then tar up the genimage and copy the tar file back to the nfs mounted directory, and then change the osimage attribute rootimgdir back to the nfs mounted directory.

2.9 Release Notes


New Function and Changes in Behavior

  • Ubuntu 14.04 and 14.04.01 support on x86_64 and ppc64le platforms
    • diskful and diskless
    • kit support
    • Statelite is not support
    • Hierarchy is not support
  • Ubuntu 14.10 support(experimental)
  • SLES 12 support on x86_64 and ppc64le platforms
    • SLES 12 diskful provisioning
    • Running xCAT management node on SLES 12
    • SLES 12 ppc64le could only run on PowerKVM virtual machines
    • Diskless is not supported
    • Hierarchy is not supported
  • RHEL 6.6 support on x86_64 and ppc64
  • AIX 7.1.3.15 and AIX 7.1.3.30 support
    • Diskful rte installation
    • Hierarchy configuration
    • No diskless support
    • No DFM support for Power 8 servers
  • Power 8 LE support
    • Hardware discovery for Power 8 LE CECs
    • Hardware control for the Power 8 LE CEC through ipmi
    • Firmware update support
    • Ubuntu 14.04 and Ubuntu 14.04.01 PowerNV support
    • Infiniband FDR support with Ubuntu PowerNV
    • Running xCAT management node on Ubuntu ppc64le
    • PowerKVM hypervisor provisioning
    • PowerKVM virtual machines management
    • create, modify, list, delete
    • Virtual machines hardware management through libvirt
    • Deploy Ubuntu ppc64le and SLES 12 ppc64le onto PowerKVM virtual machines
  • sysclone enhancements
    • ppc64 RHEL 6.x and ppc64 RHEL 7 support
    • ppc64 SLES 11 SPx support
    • x86_64 RHEL 7 support
    • LVM support on RHEL
  • A new site attribute auditnosyslog is added to control if the commands written to auditlog will also be written to syslog. This attribute in combination with auditskipcmds="ALL" can turn off all logging of commands.
  • A new site attribute nmapoptions is added to specify additional options for the nmap command used in pping, nodestat, xdsh -v and updatenode commands. See the "tabdump -d site" for more details.
  • Putting the user customized postscripts into subdirectories under /install/postscripts/ is supported, see the doc Adding your own postscripts for more details.
  • The following features are dropped and no longer supported in xCAT 2.9:
    • RHEL 5.x and CentOS 5.x support
    • The whole Fedora support is dropped
    • SLES 10 SPx support
    • Ubuntu 12.04, 12.10, 13.04, 13.10 support
    • AIX diskless deployment is only supported on Power 775 clusters.
    • vSphere 4.1 support
    • z/VM 5.1, 6.1 support
    • IBM Bladecenter support
    • xCAT web interface(xCAT-UI)
    • Monitoring through IBM Resource Monitoring Control(xCAT-rmc)
    • xCAT OpenStack support (xCAT-OpenStack and xCAT-OpenStack-baremetal)
    • xCAT IBM HPC integration is done through kits, drop the package xCAT-IBMhpc

Test Environment

Key Bug fixes

Restrictions and Known Problems

  • The perl-Net-DNS-0.73-1.28 shipped with SLES 12 has one bug that causes makedns could not work, see xCAT bug 4331 for more details. We are working with SuSE to update the perl-Net-DNS package in SLES, as a temporary solution, xCAT compiled and shipped an updated version of perl-Net-DNS that fixed the known problem in xcat-dep, the Perl-Net-DNS package shipped with xcat-dep is perl-Net-DNS-0.80-1, you need to update the perl-Net-DNS to the version shipped with xcat-dep manually, if the update was not done.
  • On stateless x86_64 Ubuntu 14.04.01 compute nodes, "cat /etc/*release*" lists Ubuntu 14.04, the workaround is to check the kernel version to verify if the statless node is Ubuntu 14.04 or 14.04.01.
  • DFM on AIX does not support IBM Power 8 machines.
  • In DFM configuration, the following rspconfig options are not working when site.enableASMI is set to yes: memdecfg decfg procdecfg iocap time date autopower sysdump spdump network dev celogin1. These options need to go through ASMI web interface, there are some major problems in the code logic to support the ASMI web interface.
  • If there are x86_64 KVM or PowerKVM guests in the xCAT cluster, the perl-Sys-Virt is required to be installed manually on the management node to enable hardware control capabilities against the virtual machines. See bug 4344 for more details.
  • kdump on rhels7 ramdisk-based statelite is not supported
  • For Ubuntu 14.04 and 14.10, the diskful provisioning and diskless genimage will need internet connections to download additional packages from internet Ubuntu repo, this is definite requirement for now, we will investigate if it is possible to eliminate this requirement in some later release.
  • Redhat 7 deployment on Power 7 platform sometimes fail with "error: connection timeout.", the workaround is to change the linux kernel and initrd download method from "http" to "tftp" in grub2. See 4406 for more details.
  • If there are quite a few(e.g. 12) network adapters on the SLES compute nodes, the os provisioning might hang because that the kernel would timeout waiting for the network driver to initialize. The symptom is the compute node could not find os provisioning repository, the error message is "Please make sure your installation medium is available. Retry?". See bug 4462 for more details.

To avoid this problem, you could specify the kernel parameter "netwait" to have the kernel wait the network adapters initialization. On a node with 12 network adapters, the netwait=60 did the trick.

chdef <nodename> -p addkcmdline="netwait=60"
  • The xCAT postscripts may not be able to run when provisioning SLES 12 on x86_64 platform, the symptom is that after the os provisioning is done, the node status is still "booting", and the node could not be logged in through ssh. The workaround is to login the node through console and run "service sshd start", and then run xdsh <nodename> -K", and then "updatenode <nodename> -P". See bug 4463 for more details.

  • On Ubuntu 14.x, the command 'lsdef <node> -i status' sometimes can NOT return the correct value for the status attribute. This issue also impacts the attributes which are stored in the nodelist table. Refer to the bug 4468 for more details.

  • The Postscripts otherpkgs didn't support on sles12. Refer to the bug 4482 for more details.

2.8.5 Release Notes


New Function and Changes in Behavior

  • The xCAT documentation pages have been migrated from the Mediawiki to the sourceforge Allura wiki, see xCAT main page and xCAT documentation page for more details.
  • IBM Power 8 servers in PowerVM mode are now supported.
  • rhels7 is supported on ppc64 and x86_64.
  • For rhels7 on ppc64 xCAT uses grub2 instead of yaboot to deploy ppc64 nodes. yaboot was deprecated in rhels7.
  • rhels7 by default uses the consistent and predictable network device naming for network interfaces. These features change the name of network interfaces from traditional "eth[0...9]" to predictable network device names, for example: enp96s0fx. See CONSISTENT NETWORK DEVICE NAMING for more details. To use the traditional network interfaces naming mechanism, add the "net.ifnames=0" to "addkcmdline" attribute of node or osimage.
  • ubuntu 14.04 is supported on x86_64 with some known problem listed in the "Restrictions and Known Problems" section.
  • MySQL has been dropped in rhels7. The mysqlscript will install and setup MariaDB on rhels7 if the rpms are installed.
  • Add support in xcatconfig -s, remoteshell postscript to support the generation and distribution of ssh hostkey "ssh_host_ecdsa_key" that is now generated with newer releases of openssh.
  • Sysclone support update delta changes in sles, partly support in redhat and centos (please refer to Restrictions and Known Problems).
  • Added postscript configbond which can be used to configure bond device on the compute node.
  • Added the replace operation support in postscript routeop. Changed the postscript setroute which is now using 'roupteop replace' as the default operation.

Test Environment

Key Bug fixes

  • mknb runs twice during installation/upgrade of xCAT bug 4150
  • xdsh -K supports vios node bug 4142
  • per node consoleondemand support bug 4136
  • mkdef/chdef does not handle the nodegroup correctly for nic* attributes bug 4130
  • All the 2.8.5 bug fixes 2.8.5 fixes

Restrictions and Known Problems

  • rhels6.4 NFS based statelite is currently unusable on both ppc64 and x86_64 due to a rhels6.4 kernel bug: bug 3535. This bug has been fixed in kernel-2.6.32-431.el6.ppc64.rpm shipped in rhels6.5.
  • renergy does not support Power 8.
  • Sysclone update delta changes has limitation in redhat and centos. when your delta changes related bootloader, it would encounter error. This issue will be fixed in xcat higher version. So up to now, in redhat and centos, this feature just update files not related bootloader.
  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.
  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm.
  • ubuntu 14.04 automatic provision is broken off by some confirmation dialogs, See defect bug 4214.
  • Due to a Redhat7 kexec-tools bug, kdump on rhels7 might fail with "kdump: wrong kdumpnic: eth2. kdump: get_host_ip exited with non-zero status!",please refer to the defect bug 4080 for workaround.
  • kdump on rhels7 statelite is not supported. Refer to the defect bug 4080 for more details.
  • rhels7.0 service node otherpkgs list, refer to service.rhels6.ppc64.otherpkgs.pkglist
  • syntax error in hardeths - see defect for the fix 4270
  • failed to run genimage against CentOS and Oracle Linux. Refer to the defect bug 4279 for work around.
  • multiple xcatd DB Access threads being left around. Refer to the defect bug 4284
  • If does not set installnic (it's null) and the real installnic is not eth0, the OS deployment might hang. Refer to the defect bug 4297
  • The OS deployment will fail if the mac attribute of your target node looks like this: 6c:ae:8b:3c:a8:ca|6c:ae:8b:3c:a8:cb!noip|6c:ae:8b:3c:a8:cc!noip . Refer to the defect bug 4299

2.8.4 Release Notes


Overview

This page documents all of the changes and notes for

 xCAT 2.8.4

Released on

 May 23, 2014

A combined summary of all recent xCAT release notes can be viewed at XCAT Release Notes

Notice: We repackaged xCAT 2.8.4 on May 29, 2014 for defect 4145. The code is the same. We only added a file /etc/httpd/conf.d/xcat.conf into xCAT-server package.

New Function and Changes in Behavior

  • rhels6.5 and rhels5.10 are now supported
  • AIX 7100-03-02 support
  • xCAT cluster zones, see documentation for details Setting_Up_Zones.
  • tabprune -a can be used on any xCAT table to remove all entries in the table.
  • The xCAT version from lsxcatd -v and the other commands now includes the git commit information. For example: Version 2.8.4 (git commit bb06e4479e68e71723c4c4769fb0837304c90a0e, built Wed Apr 2 05:28:30 EDT 2014)
  • xCAT OpenStack baremetal driver, see documentation for details [Using_xCAT_in_OpenStack_Baremetal_Node_Deployment]
  • makedns master/slave support, see documentation for details [Cluster_Name_Resolution]
  • A new flag '--ignorekernelchk' for commands genimage,geninitrd and nodeset to skip the kernel version checking when injecting drivers from driver rpm to initrd.
  • Windows support enhancements:
    • Secondary adapters support
    • Multiple partitions support
    • Multiple WinPE
    • Documentation updates
    • postscript support
  • DFM enhancements:
  • Support static network configuration during node provisioning, can be enabled by setting "site.managedaddressmode=static". Currently, this feature is only available for diskfull installation of redhat and sles.
  • REST-API restructure(experimental). See documentation for details [WS_API]
  • Ubuntu 14.04 diskful installation support (experimental)
  • Statelite enhancement that 'persistent' directory will mount to the node specific directory on the nfs server from the compute node so that one node can NOT see the file/dirs of other nodes in the persistent mount directory.

Test Environment

The following list summarizes the specific details of the test environments used for this release of xCAT. Although not specifically tested, xCAT will continue to be supported and work correctly for other environments. We just do not have the resources to test all possible operating systems and hardware for each release.

For a combined summary of all test environments for recent releases of xCAT, see: XCAT Test environment Summary

Key Bug fixes

  • xdsh -e -E not working correctly in hierarchical environment. bug 4112
  • xdcp rsync issues with (merge,append, execute, executealways) in hierarchical environment. bug 4061
  • xcatd restart on service node hits invalid error check when it runs nodeset on the servicenode. bug 3942
  • Add vlan support for configeth. bug 4025
  • makedhcp -n to add bridges into dhcpd.conf bug 3902
  • Sysclone enhancements to support golden client with extended and logical partitions bug 3940
  • node reinstall loop when site.nodestatus=0 bug 3997
  • stateless and statelite netboot failed for kvm virtual machine bug 4096
  • Additional 2.8.4 fixed defects 2.8.4 bugs

Restrictions and Known Problems

  • rhels6.4 NFS based statelite is currently unusable on both ppc64 and x86_64 due to a rhels6.4 kernel bug: bug 3535. This bug has been fixed in kernel-2.6.32-431.el6.ppc64.rpm shipped in rhels6.5.

Upgrade of 2.7 to 2.8 on SLES

  • When updating xCAT from 2.7 to 2.8 on a SLES x86_64 MN, using the command zypper update -t package 'xCAT' , zypper will ask the following question:

    Problem: xCAT-2.8-snap201302071009.x86_64 requires xCAT-genesis-scripts-x86_64, but this requirement cannot be provided
    uninstallable providers: xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch[xcat28]
    Solution 1: replacement of xCAT-genesis-x86_64-1:2.7.7-snap201301100842.noarch with xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch
    Solution 2: do not install xCAT-2.8-snap201302071009.x86_64
    Solution 3: break xCAT by ignoring some of its dependencies

    Choose from above solutions by number or cancel [1/2/3/c] (c):

  • Choose solution # 1. The new xCAT-genesis-scripts-x86_64 rpm replaces the xCAT-genesis-x86_64 rpm. It gets combined with the xCAT-genesis-base-x86_64 rpm from xcat-dep when mknb is run and forms the new genesis boot kernel.

  • If you have a hierarchical SLES x86_64 cluster, you will hit this same problem when upgrading the services nodes. Since the choice needs to be responded to interactively, you must upgrade xCAT on the SNs manually, instead of having xCAT's otherpkgs support do it for you.

  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.

  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.
  • When "site.managedaddressmode=static", sles provisioning on system X might hang, the details and workaround can be found in bug #4132.
  • on ubuntu MN, "lsxcatd -v" and "-v" option of other xcat commands failed to get xcat release info. See bug #4128

2.8.3 Release Notes


New Function and Changes in Behavior

  • The policy table is now sorted on the priority field before checking it for authorization for a particular command. See bug 2959.
  • New support for rebuilding the initrd of an OS image to inject new drivers from rpm packages. See [Using_Linux_Driver_Update_Disk].
    • Added a new command 'geninitrd <osimage>' to rebuild the initrd for both stateful and stateless osimages.
    • Added a new flag 'nodeset --noupdateinitrd' to avoid the rebuild of the initrd for a stateful osimage (if you have already done it using geninitrd).
    • Added the function to search the 'osimage.osupdatename' to get the kernel rpm and extract the new drivers to the initrd from the searched kernel rpm. If the new kernel is included in the 'osimage.osupdatename', it will be used in the os deployment process for this osimage.
  • Xeon Phi Support Phase 2: See [Managing_MIC_(Intel_Xeon_Phi)_nodes]
    • Support MPSS 3.1 (Built with Yocto). (The MPSS 2.x support has been dropped from 2.8.3)
    • Support software installing for mic node. Three types of format are supported for customer to install software on mic node:
    • filelist format - All the files will be installed to a chroot directory. A specific .filelist configuration file needs be installed at /opt/mic to indicate which file should be installed to where in mic ramfs.
    • rpm format - The rpm will be copied to ramfs of mic node, it will be installed just before the running of init during boot of mic Linux system.
    • simple format - A directory is specified that the whole directory will be copied directly to mic ramfs.
    • Support to use stateless OS on mic host node. The MPSS is installed by kit to the stateless image for mic host. After adding kit to osimage, the MPSS will be installed automatically to chroot directory when running genimage against the stateless image.
    • Support auto nfs mount during mic booting
    • Support internal bridge
  • NeXtScale support: See [XCAT_NeXtScale_Clusters]
    • Support for discovery of the NeXtScale Fan Power Controllers (FPC)
    • rpower, rinv, rflash, and rvitals support for the NeXtScale FPC
    • Support for discovery, deployment, and managment of the IBM NeXtScale nx360 M4 Compute Node.
  • Deploying OpenStack support on ubuntu 12.04: See Deploying OpenStack with Chef
    • Update the clouds table
    • Update the scripts: confignics, config_chef_server, config_chef_workstation and config_chef_client ,
    • Add a new command makeclouddata to generate the environment file
    • Add new scripts: configbr-ex(using with confignics), mountinstall, loadclouddata, configgw
  • Remove machine type check for renergy command
    • Enhance the renergy command for System P that renergy command can be run against any power machine node.
  • xdsh now supports the -t flag for timeout when running ssh key updates (-K). See xdsh man page. updatenode also supports the -t flag for the similar (-k) option.
  • tabrestore new flag -a will add the rows from the .csv file to the table instead of replacing all table contents from the .csv file.
  • xCAT flow control - performance improvement for deploying large clusters. See Using Flow Control for more details.
  • The group ownership of the files in the xcat-core and xcat-dep tarballs was changed from "xcat" to "root". Also, the permissions of the files in the xcat-dep tarball were cleaned up to make them consistent and more appropriate.
  • Imaged provisioning for sles11, sles10, rhel6, rhel5 on system X. See section Installing Stateful Nodes Using Sysclone.
  • Confignics enhancements:
  • Sequential discovery enhancements
    • Added a new flag "-n" to nodediscoverstart, to specify to run makedns <nodename> for any new discovered node
    • Added a new argument osimage=xxx to nodediscoverstart, to specify the discovered nodes will be associated with the osimage and the os provisioning should be started automatically.
    • If the bmciprange is specified with nodediscoverstart, it will set up the BMC for any new discovered nodes automatically during the sequential discovery process
    • A new flag "-s|--skipbmcsetup" is added to skip the bmcsetup even when bmciprange is specified
  • Support AIX 7.1.3.0
    • There are two versions of perl-Net_SSLeay.pm rpm listed in the sample bundle files, use perl-Net_SSLeay.pm-1.30-3 for AIX 7.1.2 and older versions, use perl-Net_SSLeay.pm-1.55-3 for AIX 7.1.3 and above, see details in /opt/xcat/share/xcat/installp_bundles/xCATaixCN71.bnd and /opt/xcat/share/xcat/installp_bundles/xCATaixSN71.bnd.
    • Also see the known problem below about the net-snmp RPMs in bundle files
  • Enhancements to Kit support
    • Allow multiple package locations as input to the buildkit addpkgs command.
    • Support updating packages in a complete Kit.

Test Environment

  • Operating systems verified with this xCAT release:
    • AIX: 7.1.2, 7.3.1.0 and 7.3.1.1 ( 71L and 71L sp1)
    • LInux: Rhel5,6 SLES 10,11

Key Bug fixes

  • Check for Management node in the database did not work for some MN names. bug 3778
  • xdsh commands timeout to BNT switch bug 3777
  • Incorrect password chosen for r* command. See bug 3780
  • remoteshell postscript can infinite loop on error . See bug 3781
  • tabprune -d does not work on eventlog bug 3823
  • imgexport 2.7 will not imgimport to 2.8 bug 3813
  • packimage (squashfs) option not working bug 3683
  • For other closed 2.8.3 defects 2.8.3 defects

Restrictions and Known Problems

  • If there is more than one nic on the management node or service nodes that is configured with ip addresses in the same subnet, only the first nic in this subnet can be used as the dhcpinterface. The other nics in this subnet will be ignored by xCAT DHCP setup. See SF defect bug 3792 for more details.
  • rhels6.4 NFS based statelite is currently unusable due to a rhels6.4 bug: bug 3559. There is no work around yet. This problem is fixed in rhel 6.5, which will be supported in xCAT 2.8.4.
  • rhels6.4 statelite will fail if noderes.xcatmaster is set to the hostname of the MN due to bug 3693. The work around is to set noderes.xcatmaster to the ip address of the MN.
  • For deploying OpenStack support on ubuntu 12.04, the configbr-ex couldn't work well. You can use the e-fix to replace the file /install/postscripts/configbr-ex. See SF defect bug 3898 for more details.
  • For deploying OpenStack support on ubuntu 12.04, if there isn't an ./environments/ in /install/chef-cookbooks, the makeclouddata couldn't work well. See SF defect bug 3904 for more details.
  • Uninstalling PE 1.3 PTF6 man page ppe_rte_man-1.3.0.6-*.rpm either directly or indirectly by upgrading to PE 1.3 PTF7 will output RPMTransaction errors during the rpm %preun script processing. The new packages are installed successfully (despite the errors). The errors can be ignored (we think), so no workaround is required. Waiting for a fix from PE RTE. See SF defect bug 3486 for more details.
  • On a new installation of xCAT 2.8.3 on an AIX or SLES 11 SP3 ppc64 management node, if you are using mysqlsetup you may get an error that it can't connect to xcatd. The work around is to simply run mysqlsetup again and it should succeed the second time. See bug 3906 for details.
  • If you have AIX 7.1.3 (or above) service nodes, you must update the bundle file that is used to install xCAT to install the later version of of the net-snmp-* RPMs that are now in the xcat-dep tarball. The bundle file that is shipped with xCAT, that can be used as a sample for your service node bundle file, is in /opt/xcat/share/xcat/installp_bundles/xCATaixSN71.bnd on the management node. But for AIX 7.1.3 or above, it should install net-snmp-5.7.2-3, not net-snmp-5.4.2.1-3. See bug 3912 for details.
  • Driver update from media fails on SLES 11.3. See bug 3933.
  • In the hierarchy cluster, if the node's xcatmaster is set to the service node, nodeset <noderange> rumcmd=xxx and nodeset <noderange> shell will still point the node to use the management node as the xcatmaster. See bug bug 3932.
  • Error on service xcatd restart on Service Node "Nodeset was run with a noderange containing both service nodes and compute nodes. This is not valid. You must submit with either compute nodes in the noderange or service nodes. " Fix provide on defect. See bug 3942.

Upgrade of 2.7 to 2.8 on SLES

  • When updating xCAT from 2.7 to 2.8 on a SLES x86_64 MN, using the command zypper update -t package 'xCAT' , zypper will ask the following question:

    Problem: xCAT-2.8-snap201302071009.x86_64 requires xCAT-genesis-scripts-x86_64, but this requirement cannot be provided
    uninstallable providers: xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch[xcat28]
    Solution 1: replacement of xCAT-genesis-x86_64-1:2.7.7-snap201301100842.noarch with xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch
    Solution 2: do not install xCAT-2.8-snap201302071009.x86_64
    Solution 3: break xCAT by ignoring some of its dependencies

    Choose from above solutions by number or cancel [1/2/3/c] (c):

  • Choose solution # 1. The new xCAT-genesis-scripts-x86_64 rpm replaces the xCAT-genesis-x86_64 rpm. It gets combined with the xCAT-genesis-base-x86_64 rpm from xcat-dep when mknb is run and forms the new genesis boot kernel.

  • If you have a hierarchical SLES x86_64 cluster, you will hit this same problem when upgrading the services nodes. Since the choice needs to be responded to interactively, you must upgrade xCAT on the SNs manually, instead of having xCAT's otherpkgs support do it for you.

  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.

  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.

2.8.2 Release Notes


New Function and Changes in Behavior

  • site.precreatemypostscripts attribute is now supported on AIX
  • xdsh now sets environment variable NODE to the nodename (as known by xCAT in the database) when running the command on the node. This can be useful if the command is a script you wrote that needs to know the nodename of the node it is running on (where the hostname value could have been changed). To see the current exports, run: xdsh <nodename> -T hostname
  • Performance enhancements in updatenode -P function.
  • Node cloning, aka imaged provisioning (and we call it sysclone) is now supported on xRHEL and CentOS. (xCAT uses SystemImager for this support.) See Installing Stateful Nodes Using Sysclone
  • xCAT can now be used to quickly deploy an OpenStack cloud. See [Deploying_OpenStack]
  • Sequential discovery enhancements. See Sequential_Discovery
    • A new command called nodediscoverdef has been added for use during the process of discovering a node manually. It can also be used to clean the discovered nodes from discoverydata table.
    • The bmc name can now be any string that is set in the ipmi.bmc attribute. The previously the bmc name had to be <node>-bmc for sequential discovery.
    • The otherinterface attribute of the node can be used to specify the IP address for the bmc during discovery.
  • The option to use the node's local disk for non-state data (swap, tmp, trace, etc.) is now supported for stateless nodes. (It was previously only supported for statelite nodes.) See Using localdisk
  • Support for managing Intel Xeon phi MIC cards: install necessary software, configure networking, create images for the cards, control the cards (boot up, power off). This support is currently experimental, but should be "hardened" soon. Try it out and give us feedback. See [Managing_MIC_(Intel_Xeon_Phi)_nodes]. Some current limitations:
    • rcons does not work well
    • For the nodeset command, the osimage needs be specified explicitly and the IP/hostname entry for MN should be added to /etc/hosts in host nodes
  • There are a few predefined groups based on hw types in the templates under directory /opt/xcat/share/xcat/templates/e1350 and now in /opt/xcat/share/xcat/templates/power. If you import these tables, and add your nodes to the corresponding groups, they will automatically have some of the necessary attributes defined.
  • Kits enhancements:
    • IBM HPC kits for ppc64 (will be available later this summer)
    • More settings when building kits to be able to control the order in which the rpms are installed. This enables some kitcomponents to be installed without work arounds in more situations.
    • New buildkit command line option "-l | --kitloc <kit_location>". This allows you to specify a kit directory location other then the current directory.
    • New kit protocol settings within kits so that when a kit is added to a cluster, the xcat code can tell if the version of xcat this kit was build with is compatible with the version of xcat it is being installed in.
    • New document giving lots of information about making your own kits: [Building_Software_Kits]
  • SLES 11 SP3 support on system x and system p
  • The rinstall command now supports nodes that are using an osimage definition.
  • The [Highly_Available_Management_Node] documentation has been reorganized a little to separate the method of sharing data from the method of failing over services.
  • Ubuntu support enhancements:
    • renergy support
    • xCAT no longer changes /bin/sh to link to /bin/bash (it leaves it as a link to /bin/dash)
    • makeroutes support
    • Support for the confignics and configeth postscripts and use of the nics table
  • The mkvlan command now works with the nics table. The mkvlan command is not available in base xCAT. Currently, it is only available with the PCM-AE product.)
  • A new flag -q was added to makedhcp to query the node entries from the DHCP server configuration.

Key Bug fixes

  • You may use FQDN for your nodes in the database. We do not have known problesm. xCAT recommends using short hostnames in the database and that is how the code is tested.
  • rinstall now supports provmethod=osimagename bug 3463.
  • instoss script on AIX will call updtvpkg internally to avoid missing library error. bug 3677
  • Fixed a problem for rhels6.4 MN installing rhels5.9 CN. bug 3598
  • makedns now handles the add/remove of the node names defined in the nics table. bug 3604
  • The perl-IO-Socket-SSL on sles10 xCAT MN needs be updated to perl-IO-Socket-SSL-1.77 in xcat-dep. bug 3699

Restrictions and Known Problems

  • rmdsklsnode can not remove the NIM machine definition with AIX 7.1.2.0 Update image on xCAT management node or service node. This is actually caused by AIX APAR IV32670. You can get more info from bug 3527

    the workaround is to force reinstall the bos.sysmgt.nim.master 7.1.2.0 fileset using AIX 7.1 TL02 media.

  • Uninstalling PE 1.3 PTF1 man page ppe_rte_man-1.3.0.1-.rpm either directly or indirectly by upgrading to PE 1.3 PTF2 fails during the rpm %preun script processing. This has been fixed in PE 1.3 PTF2 ppe_rte_man-1.3.0.2-.rpm.

    To work around this problem, you may need to directly remove the rpm using "rpm -e ". If you are working with a diskless image, the genimage command may fail the first time you try to remove PE 1.3.0.1 or try to upgrade to 1.3.0.2. Simply run your genimage command again, and it should work correctly the second time.

  • makedhcp -a does not always update the DHCP leases correctly. Workaround is to run makedhcp -n and makedhcp -a. bug 3535

  • rhels6.4 NFS based statelite is currently unusable due to a rhels6.4 bug bug 3535. There is no work around yet.
  • sysclone only works for system x RHEL 6.x and CentOS 6.x.
  • rcons might fail when nodehm.mgt=kvm, see bug 3719.
  • xlc/essl kit dependency issue. See defect for problem description and patch bug 3746 .

Upgrade of 2.7 to 2.8 on SLES

  • When updating xCAT from 2.7 to 2.8 on a SLES x86_64 MN, using the command zypper update -t package 'xCAT' , zypper will ask the following question:

    Problem: xCAT-2.8-snap201302071009.x86_64 requires xCAT-genesis-scripts-x86_64, but this requirement cannot be provided
    uninstallable providers: xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch[xcat28]
    Solution 1: replacement of xCAT-genesis-x86_64-1:2.7.7-snap201301100842.noarch with xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch
    Solution 2: do not install xCAT-2.8-snap201302071009.x86_64
    Solution 3: break xCAT by ignoring some of its dependencies

    Choose from above solutions by number or cancel [1/2/3/c] (c):

  • Choose solution # 1. The new xCAT-genesis-scripts-x86_64 rpm replaces the xCAT-genesis-x86_64 rpm. It gets combined with the xCAT-genesis-base-x86_64 rpm from xcat-dep when mknb is run and forms the new genesis boot kernel.

  • If you have a hierarchical SLES x86_64 cluster, you will hit this same problem when upgrading the services nodes. Since the choice needs to be responded to interactively, you must upgrade xCAT on the SNs manually, instead of having xCAT's otherpkgs support do it for you.

  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.

  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.

2.8.1 Release Notes


New Function and Changes in Behavior

  • Release 2.8.1 is now available for AIX.
  • xCAT 2.8.1 deprecating support of rsh/rcp for remote commands xdsh/xdcp.
  • Linux x86_64 RHEL 5 users need to apply the latest deps package and then can use 2.8.1.
  • Support multiple paths with the osimage attribute pkgdir, see [Install_OS_Updates]
  • Added image name and timestamp to file /opt/xcat/xcatinfo in diskless images
  • Energy management support for Flex nodes
  • Sequential discovery support, see [XCAT_iDataPlex_Cluster_Quick_Start#Sequential_Discovery]
  • Updates for system x Flex management process, see [XCAT_system_x_support_for_IBM_Flex]
  • Enhancements to xCAT Software Kit support:
    • release attribute for kits
    • default kit tarfile names include kit architecture
    • buildkit addpkgs allows changing kit version and/or release
    • addkitcomp --noupgrade option to allow multiple releases of a kitcomponent to exist in one osimage
    • new lskit, lskitcomp, lskitdeploymentparams commands
    • various buildkit, addkit, addkitcomp defect fixes
  • Statefull images creation for management node
  • Kits installation on management node
  • Procedure on how to convert non-osimage based system to osimage based system, see [Convert_Non-osimage_Based_System_To_Osimage_Based_System]
  • A new flag -s to reventlog command to sort the output.
  • RHEL 6.4 support on system x and system p
  • RHEL5.9 support on system x and system p
  • IPv6 support enhancements on Linux, see [Configuring_IPv6_in_Cluster]
    • confignics support to configure IPv6 addresses on Ethernet and Infiniband interfaces
    • makeroutes support for IPv6 routes
    • makehosts and makedns support IPv6 hosts
  • lsdef,chdef,mkdef to support display/set nic attributes more easily
  • Ubuntu support enhancements
    • hardware discovery
    • mysql and postgresql support
    • makeknownhosts -r support
    • kit support
  • Support for Management Node in the servicenode table to setup attributes. See [Managing_the_Management_Node]
  • Install and config chef/puppet for OpenStack deployment(experimental). See [Adding_Chef_in_xCAT_cluster] and [Adding_Puppet_in_xCAT_cluster]
  • nicaliases support in the nics table
  • xdsh -E works in a hierarchical cluster
  • xdcp ( scp and rsync ) support sudo. updatenode -F supports sudo. Hierarchical custers are also supported. See [Granting_Users_xCAT_privileges].
  • Energy management support for Flex system (renergy command)
  • Sequential discovery support (nodediscoverstart, nodediscoverstop, nodediscoverls, nodediscoverstatus). The simplest method to discovery new hardwares for a physical location unaware cluster. Node Discovery
  • To migration from xCAT 2.7.x or earier to xCAT 2.8.1 or later, there are some additional steps need to considered:

Key Bug fixes

  • You can install xCAT 2.8.1 on RHEL 5 (ppc64 and x86_64), AIX 6.1, SLES 10. The following releases: SLES10 x86_64 or ppc64 and Redhat5 ppc64 no longer supports genesis discovery. defect 3426. Need to install latest xCAT deps package.
  • Errors returned from xdsh/xdcp now have Error:<nodename>: on the front instead of just <nodename>:. This will break the use of xdshbak or xcoll to sort. defect 3380.
  • The ppping command will return a usage error if you use the -i flag. For now, use the --interface flag instead. See bug 3386.
  • The genimage -l flag does not work correctly to limit the root filesystem on RHEL 6. See bug 2972.
  • The xdsh -E flag does not work in a hierarchical cluster (i.e. one with service nodes). See bug 3052.
  • The python RPM is also needed as a prereq for the Mellanox IB driver on SLES 11.2. Use the updated ib.sles11.2.x86_64.pkglist in bug 3350.
  • If you have Linux nodes with FQDN hostname, you will find that the running of postscripts (e.g. updatenode -P) will fail. 3398
  • imgcapture fixed 3436.
  • Incorrect postscript/postbootscript list generated for a node. 3412
  • sles11.2 nfs_based statelite on x86 deployment now works. See SF bug 3038.
  • The rcons command works for a node whose noderes.conserver attribute is explicitly set to the management node, see 3159
  • genimage -l works with RHEL6, see see 2972
  • New xCAT SSL certificates not working for hierarchical commands - 3507

Restrictions and Known Problems

  • Cannot use fully qualified hostnames in the xCAT database.
  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES11 SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.
  • Uninstalling PE 1.3 PTF1 man page ppe_rte_man-1.3.0.1-.rpm either directly or indirectly by upgrading to PE 1.3 PTF2 fails during the rpm %preun script processing. This has been fixed in PE 1.3 PTF2 ppe_rte_man-1.3.0.2-.rpm.

    To work around this problem, you may need to directly remove the rpm using "rpm -e ". If you are working with a diskless image, the genimage command may fail the first time you try to remove PE 1.3.0.1 or try to upgrade to 1.3.0.2. Simply run your genimage command again, and it should work correctly the second time.

  • makedhcp -a does not always update the DHCP leases correctly. Workaround is to run makedhcp -n and makedhcp -a. bug 3535

  • rinstall does not support provmethod=osimagename bug 3463.
  • List of defects to be fixed in 2.8.2 defects
  • The policy table does not have an entry for remoteimmsetup, if xCAT is upgraded from a previous version to 2.8.x. Adding the following line in the policy table addresses this: "2.1",,,"remoteimmsetup",,,,"allow",, bug 3554.
  • Certain NodeRanges do not expand correctly, see bug 3429.
  • The linuximage table entries, when created automatically by xCAT, may contain the wrong template information, as the matches to the existing template files aren't done specifically enough.
  • The rspconfig <noderange> textid command can return "No name" if a blade or ITE is defined with a slot in the mp table in which there is no IMM connection to the AMM or CMM. In this case, even if the blade or ITE takes up more than one bay (e.g., a Flex x440, or a blade with a BGE), the mp table entry for that blade or ITE should list only one slot. The case where two slots would be defined is a double-wide HX5 blade, where there is an IMM connected in each slot.
  • For Redhat 6.4 NFS based statelite, the tmpfs files defined in litefile table are still readonly, See bug 3559.
  • Cannot install xCAT 2.8 on any version before rhel5 and sles10. You can install rhels5 (x86_64), if the latest xCAT deps package is installed. New node discovery on X-series will not work for SLES 10 or earlier releases of SLES . See bug 3426.
  • servicenode table entry, with no service nodes defined causes problems in xCAT commands. bug 3580
  • xcat client could not connect to xcatd through IPv6 link local address. bug 3581

Upgrade of 2.7 to 2.8 on SLES

  • When updating xCAT from 2.7 to 2.8 on a SLES x86_64 MN, using the command zypper update -t package 'xCAT' , zypper will ask the following question:

    Problem: xCAT-2.8-snap201302071009.x86_64 requires xCAT-genesis-scripts-x86_64, but this requirement cannot be provided
    uninstallable providers: xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch[xcat28]
    Solution 1: replacement of xCAT-genesis-x86_64-1:2.7.7-snap201301100842.noarch with xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch
    Solution 2: do not install xCAT-2.8-snap201302071009.x86_64
    Solution 3: break xCAT by ignoring some of its dependencies

    Choose from above solutions by number or cancel [1/2/3/c] (c):

  • Choose solution # 1. The new xCAT-genesis-scripts-x86_64 rpm replaces the xCAT-genesis-x86_64 rpm. It gets combined with the xCAT-genesis-base-x86_64 rpm from xcat-dep when mknb is run and forms the new genesis boot kernel.

  • If you have a hierarchical SLES x86_64 cluster, you will hit this same problem when upgrading the services nodes. Since the choice needs to be responded to interactively, you must upgrade xCAT on the SNs manually, instead of having xCAT's otherpkgs support do it for you.

  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.

  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.

2.8 Release Notes


New Function and Changes in Behavior

  • AIX users should wait for xCAT 2.8.1 before upgrading to this release.
  • Linux Redhat el5 users must wait for xCAT 2.8.1 before upgrading.
  • On Linux, after the upgrade to 2.8, you may find the xcatd daemon did not stop and restart successfully. A command like xdsh <node> date will return an error like the following:

    Can't locate object method "determinehostname" via package "xCAT::NetworkUtils" at /opt/xcat/lib/perl/xCAT_plugin/xdsh.pm line 121.

    If this occurs, run service xcatd stop and check to make sure all xcatd processes are gone. If not kill them and then run service xcatd start. This issue has been fixed in 2.8, so updates after this should not have the error.

  • Use of kits to package software so it can easily be installed in a cluster. See [Using_Software_Kits_in_OS_Images] and [IBM_HPC_Software_Kits]. Kits for IBM HPC software will be available starting in March.

  • The options "install", "netboot", and "statelite" with nodeset command are deprecated, the osimage provisioning option should be used instead. In xCAT 2.8, nodeset <noderange> install/netboot/statelite will continue to work with some warning messages, but in a future release, the nodeset <noderange> install/netboot/statelite might not work any more. For os provisioning for ESX/ESXi, RHEV-H, Windows and zVM nodes, the nodeset <noderange> install/netboot/statelite should still be used. See [Convert_Non-osimage_Based_System_To_Osimage_Based_System] for more details on how to convert non-osimage based system to osimage based system.
  • The use of bind as dns handler in the site table dnshandler attribute is deprecated. Only ddns handler has been tested. ddns has been the default in xCAT since 2.6.x. See site dnshandler attribute.
  • Added the ability to make use of a local scratch disk on statelite nodes. See [XCAT_Linux_Statelite#To_enable_the_localdisk_option].
  • Added a new osimage attribute called groups that can be used in the litefile and litetree tables instead of a single osimage name.
  • The aixremoteshell postscript will no longer appear in the postscripts default list. It is replaced by remoteshell which will be used on AIX and Linux. The remoteshell postscript will call aixremoteshell on AIX nodes. When you install or upgrade to 2.8, your postscript list will be automatically fixed.
  • The Management node can be a managed node in the database. Check this document for more details Managing the Management Node].
  • Removed support for updatenode <switch> -k --userid --devicetype. The function currently supported by xdsh is sufficient. See the xdsh man page.
  • Added a new attribute to the site table: auditskipcmds. It specifies with commands or client requests should not be logged to the audit log.
  • Enhancements to running postscripts both during install and from updatenode:
    • Setting the new site table attribute precreatemypostscripts can make postscripts run faster and put less load on xcatd in large deployments (Linux only)
    • You can now have xCAT provide additional database attributes to your postscripts by modifying the mypostscript template
    • For details on both of these enhancements, see [Postscripts_and_Prescripts] Section on Using_the_mypostscript_template.
  • New function to allow you to customize information passed to postscript and postbootscript. See the following documentation, "Using the mypostscript template": [Postscripts_and_Prescripts]
  • A new command called pasu to query or set ASU (uEFI) settings on many x86_64 nodes in parallel. See also: [XCAT_iDataPlex_Advanced_Setup] .
  • New site table runbootscripts attribute will cause postbootscripts to run on reboot of stateful (diskful ) nodes.
  • The instxcat script on AIX no longer install the xCAT-rmc-* rpm by default. This rpm has not changed in 2.8 except the version number, so if you have one installed you can continue to use it.
  • There are new node status attributes: updatestatus and updatestatustime. When updatenode is run, updatestatus will be set to "synced" or "out-of-sync" based on the success of the updatenode operation. The time of the update is recorded in updatestatustime.
  • updatenode -S is now supported to diskless nodes.
  • updatenode -l and xdsh --sudo options are added to allow you to run operations on the nodes as a non-root userid using sudo. (Currently, this is only supported in a non-hierarchical cluster.) For details, see the updatenode man page, the xdsh man page, and the setup sudo section in [Granting_Users_xCAT_privileges].
  • The default profiles for RedHat 6.x and SuSE Linux Enterprise Server 11.x now make some attempt to avoid accidental SAN install.
  • When an os image name is not specified in the command "nodeset <noderange> osimage", the os image names for the nodes will be taken from the node provmethod attribute.
  • Additional xdsh support for ethernet switches. See the xdsh man page for setup details.
  • Added support to inventory switches in sinv.
  • xCAT now supports the use of multiple hostname domains within an xCAT cluster. See Cluster Name Resolution for more details.
  • Enhanced support for specifying additional network interfaces for cluster nodes, and having xCAT automatically configure them when deploying the nodes. See Cluster Name Resolution for more details.
  • Support x86 IBM Flex systems using 'runcmd=bmcsetup' to set up the IMMs.
  • Added the node status update when the OS deployment performed through chain mechanism. See Automatically deploying nodes after discover in [XCAT_iDataPlex_Advanced_Setup]
  • Supported to extract drivers from specified rpm packages for the diskfull and diskless deployment. See Driver RPM Package in [Using_Linux_Driver_Update_Disk]
  • Changed the deployment resource from /tftpboot/xcat/netboot(install)/<os>/<arch>/<profile> to /tftpboot/xcat/osimage/<osimage name>.
  • No longer require 'ipmi' credentials to be specified for Flex, ipmi plugin now understands to use 'blade' credentials when communicating with a Flex system if ipmi not provided.
  • HA MN documentation updates
  • New support for ubuntu 12.04. The xCAT mgmt node can be run on ubuntu and both stateful and stateless ubuntu nodes can be deployed. See [Ubuntu_Quick_Start] for details. Since this is the first release of ubuntu support, you should try it in a test cluster before using it in production. Any feedback is welcome on the mailing list. Ubuntu 12.10 will be supported soon.
  • Windows Support Improvements:
    • Deprecate need for 'wintimezone' site value, the Microsoft time zone is now looked up from the POSIX value in 'timezone'
    • It is strongly suggested that genimage.cmd be used to generate a new Windows PE image.
    • New invocations of genimage.cmd with Windows ADK will add powershell to the Windows PE image.
    • Add support for Windows Server 2012, Windows 8, and Hyper-V
    • Unified template for installation with or without Active Directory
    • Support automatic fill-in of product key for retail or MAK keys changing to KMS key if no key specified.
    • Option to allow user to decline xCAT capability to pre-join domain, reducing AD requirements. However, unless very special conditions are met, administrator credentials are put at significant risk. This is controlled by the site attribute prejoinactivedirectory.
  • KVM Virtualization
    • Modify default caching scheme to be 'none' unless the storage is a cow clone of another storage device
    • Support LVM storage pools for virtual machines
  • z/VM Virtualization
    • Live migration support in z/VM 6.2 using rmigrate.
    • Improved security with no-root login in environments where root login is not allowed.
    • Added support for native SCSI/FCP devices. You can now define nodes using native SCSI/FCP devices and manage an internal storage pool for native SCSI/FCP devices.
    • Moved options for gathering storage and network configuration (e.g. diskpool, zfcppool, and network) into rinv.
    • You can now collect the inventory (e.g. number of CPUs, memory size, etc.) of any z/VM hypervisor.
    • New options are added to create networks (vSwitches and VLANs) and connect existing storage devices to the z/VM system. See chhypervisor for more info.
    • Updated zHCP (version 2.0) to support new SMAPI commands in z/VM 6.2. You can invoke any SMAPI command on the zHCP using smcli.
  • RHEV Virtualization

Key Bug fixes

  • xdcp was not handling servicenode pools correctly, see SF defect 3267:https://sourceforge.net/p/xcat/bugs/3267/
  • xcatd not stopping successfully due to pid being stored in /tmp/ and getting deleted. Changed this design and fix with defect 2966:https://sourceforge.net/p/xcat/bugs/2966/
  • updatenode -k to nodes serviced by a service node was not behaving properly. https://sourceforge.net/p/xcat/bugs/2950/
  • Use of '/' delimited regular expressions no longer produces incorrect values for unrelated fields
  • KVM directory storage URIs no longer fail with a trailing '/' character
  • KVM no longer attempts to delete .iso files backing a virtual optical drive when purging disks for rmvm -p

Restrictions and Known Problems

Most of these bugs will be fixed in 2.8.1.

  • Cannot install xCAT 2.8 on any version before rhel5. Cannot install rhels5 (ppc64). You can install rhels5 (x86_64), if the latest xCAT deps package is installed. Cannot install 2.8 on AIX 6.1 or SLES 10 or earlier releases of SLES . See defect 3426
  • Cannot use fully qualified hostnames in the xCAT database.
  • The new xCAT support for multiple network domains does not include multiple domains within a specific network. It is restricted to one domain per network.
  • Errors returned from xdsh/xdcp now have Error:<nodename>: on the front instead of just <nodename>:. This will break the use of xdshbak or xcoll to sort. defect 3380.
  • The ppping command will return a usage error if you use the -i flag. For now, use the --interface flag instead. See bug 3386.
  • The genimage -l flag does not work correctly to limit the root filesystem on RHEL 6. See bug 2972.
  • The xdsh -E flag does not work in a hierarchical cluster (i.e. one with service nodes). See bug 3052.
  • The rcons command will not work for a node whose noderes.conserver attribute is explicitly set to the management node. As a work around, leave the noderes.conserver attribute blank. See bug 3159.
  • The python RPM is also needed as a prereq for the Mellanox IB driver on SLES 11.2. Use the updated ib.sles11.2.x86_64.pkglist in bug 3350.
  • If you have Linux nodes with FQDN hostname, you will find that the running of postscripts (e.g. updatenode -P) to these nodes will fail. You will need the fix provide attached to 3398
  • imgcapture is broken. Patch available on the defect. 3436.

Upgrade of 2.7 to 2.8 on SLES

  • When updating xCAT from 2.7 to 2.8 on a SLES x86_64 MN, using the command zypper update -t package 'xCAT' , zypper will ask the following question:

    Problem: xCAT-2.8-snap201302071009.x86_64 requires xCAT-genesis-scripts-x86_64, but this requirement cannot be provided
    uninstallable providers: xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch[xcat28]
    Solution 1: replacement of xCAT-genesis-x86_64-1:2.7.7-snap201301100842.noarch with xCAT-genesis-scripts-x86_64-1:2.8-snap201302071009.noarch
    Solution 2: do not install xCAT-2.8-snap201302071009.x86_64
    Solution 3: break xCAT by ignoring some of its dependencies

    Choose from above solutions by number or cancel [1/2/3/c] (c):

  • Choose solution # 1. The new xCAT-genesis-scripts-x86_64 rpm replaces the xCAT-genesis-x86_64 rpm. It gets combined with the xCAT-genesis-base-x86_64 rpm from xcat-dep when mknb is run and forms the new genesis boot kernel.

  • If you have a hierarchical SLES x86_64 cluster, you will hit this same problem when upgrading the services nodes. Since the choice needs to be responded to interactively, you must upgrade xCAT on the SNs manually, instead of having xCAT's otherpkgs support do it for you.

  • sles11.2 nfs_based statelite on x86 deployment fails. See SF bug 3038 for workaround.

  • Perl errors on sles when perl-IO-Socket-INET6 rpm is installed on SLES SP2. See SF defects: bug 3173. You only see the messages in some commands when not running under the daemon, like xcatconfig and if you export XCATBYPASS=y and run tabdump <tablename>. To get rid of the warnings remove the rpm. Cause of problem under investigation.

2.7.8 Release Notes


New Function and Changes in Behavior

Tested OS

AIX: 7.3.1.0 and 7.3.1.1) ( 71L and 71L sp1) and AIX 6.1.9.1

LInux: Rhel5,6 SLES 10,11

Restrictions and Known Problems

  • NFS based statelite cannot configure bond0 in confighfi postscript. 3939

Key Bug fixes

  • rinstall did not support provmethod=osimagename. This support has been added. 3644
  • rnetboot hang, P5 AIX 6.1 3533
  • litefile error with mkdsklsnode -r 3183
  • confighfi postscript errors for hfi bonding 3179
  • mkhwconn -s fails for Firebird from Redhat MN 3583
  • lsslp unicast on Power Linux 3947

Restrictions and Known Problems

  • lsslp unicast doesn't support AIX.

2.7.7 Release Notes


New Function and Changes in Behavior

  • Add support for devices ( switches) in sinv. See man sinv: sinv
  • rbeacon is supported for Flex system
  • Energy management support for Flex System (renergy command)
  • RHEL 6.4 support on system x and system p

Tested OS's

RH6.3, RH6.4 and AIX71H

Key Bug fixes

  • When upgrading the xcatd did not restart properly: 2359
  • updatenode with ospkgs failure: 3229
  • LCDs were not displayed properly by the rvitals command for Flex Power 7 blades: 3382
  • In the ospkgs postscript (used by updatenode -S), the baseurl was set incorrectly in the zypper repo on SLES: 3381
  • Incorrect postscript/postbootscript list generated for a node. 3412
  • mlnxofed_ib_install need to support hierarchical environment.3513
  • Certain NodeRanges do not expand correctly, see bug 3429.

Restrictions and Known Problems

  • rmdsklsnode can not remove the NIM machine definition with AIX 7.1.2.0 Update image on xCAT management node or service node. This is actually caused by AIX APAR IV32670. You can get more info from bug 3527

    the workaround is to force reinstall the bos.sysmgt.nim.master 7.1.2.0 fileset using AIX 7.1 TL02 media.

  • updatenode -k does not work to compute nodes in a hierarchical environment; that is when the compute node is accessed via a service code. Running this command to the compute node results in an infinite loop which consumes memory. For a work around to exchange ssh keys use xdsh -K to the compute node. 3652

  • In HAMN environments, credential validation can fail. 3704

2.7.6 Release Notes


New Function and Changes in Behavior

  • HPC integration support for latest IBM HPC software stack
  • Support for SLES 10 SP4

Tested OS

  • AIX 6.1.8 and 7.1.2
  • Rhel5,6 SLES 10,11

Key Bug fixes

  • A couple IPMI bugs existed in 2.7.5:

    • In an initial install of 2.7.5 on new hardware, rvitals and rinv would not work.
    • For cmds like rvitals for a lot of nodes would give errors like: node1: Error: 1 code on opening RMCP+ session

    Both of the problems are fixed in this release, see bug 3156.

Restrictions and Known Problems

  • On Linux, if after upgrade and you run lsxcatd -a, it does not show you are running Version 2.7.6, then you will need to do the following:
    • service xcatd stop
    • ps -ef | grep xcatd, kill -9 any processes
    • service xcatd start
    • lsxcatd -a to check
    • This will be fixed in 2.7.7 with the following defect: bug 2359
  • rbeacon is not supported for Flex system, it will be supported in 2.7.7.
  • updatenode with ospkgs doesn't work in xCAT 2.7.6 . It has been fixed. You can get more information and e-fix from bug 3229
  • On AIX IB configuration, there is a known AIX IB node description issue that is fixed with AIX IFIX IV36529. xCAT has also made updates to our configiba post scripts which will be fixed in xCAT 2.7.7. You can get more information and e-fix from bug 3338
  • If you create your own osimage names, that is do not use the generated defaults, then the postscripts and postbootscripts attribute is not honored in your osimage. This is fixed in 2.7.7.

2.7.5 Release Notes


New Function and Changes in Behavior

  • Note in SVN the tag of the release was late. The last revision number for 2.7.5 is 14039.
  • RHEV (RedHat Enterprise Virtualization) (RHEV-m, RHEV-h) support. Including the installing of RHEV-h and management the virtual machines through RHEV-m.
  • Hardware discovery support for IBM Flex system x compute nodes
  • Firmware assisted dump for Power 775
  • RHEL 6.3 support
  • Numerous enhancements to the xCAT support for AIX High Availability Service Nodes (HASN).

Key Bug fixes

Restrictions and Known Problems

  • On system x hardware you must use the latest xcat-dep tarball in https://sourceforge.net/projects/xcat/files/xcat-dep/2.x_Linux/ . (Snap date 11/7/2012 or later.) Otherwise, you will encounter a bug in xCAT-genesis in which bmcsetup does not set up userids correctly on x3755 M3.
  • A couple IPMI bugs exist in 2.7.5:

    • In an initial install of 2.7.5 on new hardware, rvitals and rinv will not work.
    • For cmds like rvitals for a lot of nodes would give errors like: node1: Error: 1 code on opening RMCP+ session

    Both of the problems are fixed in the file IPMI.pm that is attached to bug 3156.

  • For xCAT 2.7.5 with Linux, you should use the xcat-dep tarball from 6/12/2012. The is the xcat-dep tarball that has been tested with xCAT 2.7.5. (The most recent xcat-dep tarball should also work with xCAT 2.7.5, but it hasn't been tested yet.)

  • For sles11.2 nfs-based statelite deployment on x86, initrd might fail to mount the rootimg and complain the messages below:

    ...
    Setting up Statelite
    mount.nfs: Protocol not supported
    Couldn't mount dx360m3n04:/install/netboot/sles11.2/x86_64/compute on /sysroot
    Trying again in 1 seconds
    ...

    the work around can be found in bug 3038

  • When running mkdsklsnode you may, in certain cases, see the following error:

    Error: there is already one directory named "", but the entry in litefile table is set to one file, please check it
    Error: Could not complete the statelite setup.
    Error: Could not update the SPOT

If you see this error simply re-run the command.

  • In an AIX HASN environment, if you have more than 8 service nodes, the mkdsklsnode command may create an /etc/exports file entry that is not supported by AIX.

    The mkdsklsnode command updates the /etc/exports file on the service nodes with an entry that contains a list of replicas that are used by the the AIX NFSv4 support. Due to an NFS limitation the list of replicas may not exceed 8. If you have existing /etc/exports files that already contain the replication entry then it will not be modified and you DO NOT need to take any further actions. However, if you have removed the /etc/exports file or need to set up a new service node, you must manually check the /etc/xports file to make sure there are no more than 8 replicas listed.

    The format of the file is as follows:

    /install -vers=4,replicas=/install@20.10.12.1:/install@20.10.12.2:/install@20.10.12.3:/install@20.10.12.4:/install@20.10.12.5:/install@20.10.12.6,noauto,rw,root=*

    Simply remove one or more replicas from the list.( ex. :/install@20.10.12.6)

  • For the rvitals command on system p with option "lcds", if there is any LPAR in the noderange which is defined in the xCAT DB, but doesn't exist on its CEC, the rvitals will not return correct info. You can get more info from bug 3133 and you also can get the e-fix to fix this issue from the attached in the bug 3133 page.

2.7.4 Release Notes


New Function and Changes in Behavior

  • SLES 11 SP2 support on System x and System p servers
  • Firmware assisted dump support for Power 775 servers
  • Support for IBM Flex system x compute nodes
  • Improved IPMI reliability and performance for large systems (~5,000 servers in 30 seconds)
  • The statelite "litetree" table now supports a mount options attribute.

Key Bug fixes

Restrictions and Known Problems

2.7.3 Release Notes


New Function and Changes in Behavior

  • pscp, prsync, psh supports new argument -f and env variable to set fanout value. See man pages.
  • SLES 11 SP2 support on system x and system p. The following features have been verified:
    • Operating system provisioning in flat (non-hierarchical) cluster, including diskful, stateless and statelite
    • Basic hardware control features
  • Support for IBM Flex system x compute nodes with RHEL 6.2.
  • HPC integration support for the latest IBM HPC software stack.
  • Deferred firmware update support for Power 775 (Experimental).

Key Bug fixes

Restrictions and Known Problems

2.7.2 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

  • Support for IBM Flex and system p compute nodes.
  • The "-t FNM" flag of rpower to boot p775 CECs from the xCAT management node, even when hierarchical DFM hardware control is configured.
  • Support for real-time service node fail over on AIX(experimental)
  • New check for xCAT ifixes. Every xCAT ifix must be removed prior to any upgrades to the xCAT packages.
  • kdump support on SLES 11
  • lsslp.pm plugin is rewritten. lsslp now uses the new SLP.pm module and does not depend on openslp.
  • HPC integration updates to install HPC software stack from otherpkg list.
  • Not display the passwords in the command output and log files
  • Scaling and performance improvements for Power 775 servers
  • Performance improvements in running postscripts, converting noderange, running synclists to large number of nodes
  • Redhat 6.2 is supported on x-series, IBM Flex System P260/P460
  • AIX 7.1 Pl1/SP3 is supported

Key Bugs Fixed

See the xCAT 2.7.2 SourceForge bugs.

Restrictions and Known Problems

  • rflash commands fails with syntax error in FSPflash.pm. See defect for affected system types. The udpate FSPflash.pm is attached to the defect. https://sourceforge.net/tracker/?func=detail&aid=3530839&group_id=208749&atid=1006945 This defect is fixed in PMR 19420. For xCAT AIX adminstrators looking for the official AIX ifix, contact IBM customer service for this PMR in the zaix,13a queue. The ifix can be downloaded from the ftp testcase.software.ibm.com download site. Note that the ifix it only remains on this download site for 5-7 days.
  • RHEL 6.2 kdump does not work on Power 775. The efix is available in this bug description.
  • lsslp multicast on AIX usage restriction. If you want to use lsslp on AIX and the nodes need to be discovered through multicast,the work around is:
    1. add multicast route first, for example: route add 239.255.255.253 40.0.0.96,
    2. use lsslp -i, for example, lsslp -i 40.0.0.96
    3. delete the route: route delete 239.255.255.253 40.0.0.96
    4. If you have several vlans and need to do discovery through several network interfaces, you need to repeat the steps above.
  • lsslp occasionally could not discover all objects: the efix has been attached on the bug https://sourceforge.net/tracker/?func=detail&aid=3529398&group_id=208749&atid=1006945
  • When installing(not updating) xCAT on system x management node, you will get some error like "Command failed: XCATBYPASS=Y /opt/xcat/sbin/mknb x86_64 2>&1. Error message: Error: Unable to find directory /opt/xcat/share/xcat/netboot/x86_64", this will cause issues with the hardware discovery for system x servers. To workaround this problem, you can either install xCATnbkernel,xCAT-nbroot* manually or make the directory /opt/xcat/share/xcat/netboot/x86_64 manually before running yum or zypper command to install xCAT.
  • genimage problem in HPC integration. There are two known problems about HPC integration. Bug: https://sourceforge.net/tracker/?func=detail&atid=1006945&aid=3529659&group_id=208749

  • 1) loadl-5103.otherpkgs.pkglist is not putting Loadleveler rpms in a subdirectory. Fix: putting the package names in loadl-5103.otherpkgs.pkglist to loadl subdirectory as:

      loadl/LoadL-full-license*
      loadl/LoadL-scheduler-full*
      loadl/LoadL-resmgr-full*
    
  • 2) GPFS otherpkgs list includes a comma in comment which is not allowed. Fix: remove the comma in GPFS otherpkgs list. For example:

      -# List additinal package names in comments to reduce image size, uncomment them if they are necessary for you.
      +# List additinal package names in comments to reduce image size
      +# uncomment them if they are necessary for you
    
  • mknb problem on system x during xCAT install, workaround provided on defect. https://sourceforge.net/tracker/?func=detail&aid=3532045&group_id=208749&atid=1006945

2.7.1 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

  • The -t flag of the genimage command has been deprecated. Use a postinstall script to replace/modify the /etc/fstab files as you like.
  • Add interactive mode for genimage command, the new flag is --interactive.
  • HPC integration updates for new HPC product versions & packaging.
  • SLES 11 SP2 support (experimental)
  • Linux postscripts logic clean up:
    • Move the creation of the mypostscript file from /tmp/mypostscript to /xcatpost/mypostscript
    • Extract the common code for generating the mypostscript file and make it common across all scripts under /opt/xcat/share/xcat/install/scripts
    • Add timestamps on the running of the postscripts and start and stop headers in /var/log/xcat/xcat.log
    • Writing xcat.log in stream like mode
  • xcatchroot command output the messages printed by the command runs in chroot environment.
  • configiba.* scripts enhancements to determine the number of IB interfaces automatically.
  • Support for installing/updating SLES SDK packages using updatenode for SLES diskful nodes.
  • Key Bugs Fixed

See the xCAT 2.7.1 SourceForge bugs.

Restrictions and Known Problems

  • SLES 11 SP2 support is currently experimental.
  • The p775 support in this release has not yet been fully tested.

2.7 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

All Environments

  • xcatd memory usage has been decreased by about two thirds
  • xCAT can now hot-load most plugin updates without a server restart.
  • xCAT now supports a site.xcatmaxconnections tunable to either restrict number of allowed SSL connections to fit in lower memory budget or increase it for more speed at the expense of memory usage.
  • xcatdebug command can be used to enable trace for xcatd and xcatd plugins without restarting the xcatd.
  • New automated test framework for xCAT.
  • New lstree command to display the tree of service node hierarchy, hardware hierarchy, or VM hierarchy
  • makedns enhancement:
    • By default, the DNS on the management node is configured. It is no longer necessary to make /etc/resolv.conf on the MN point to the DNS that makedns should configure.
    • Support new option -e to provide the flexibility to update the DNS records to an external DNS server which is listed in the /etc/resolv.conf on the management node.
    • named starts up automatically after system reboot.
  • site.ntpservers and networks.ntpservers can be set to keyword "<xcatmaster>" to specify that the management node or service node should be used for each compute node's time server.

x86_64 Hardware

  • xCAT Genesis boot image supersedes nbfs for x86_64 node discovery and other generic boot environment tasks:
    • To use it, install the xCAT-genesis-x86_64 RPM on the MN from the latest xcat-dep tarball.
    • Contains to glibc, 32 & 64 bit userspace libraries and utilities from CentOS 6.x, and allows execution of more binaries without lots of library wrangling
    • Uses CentOS 6.x kernel, and adds several drivers
    • Gathers more hardware inventory data (processors, memory, wwpn, etc)
    • Runs IBM UpdateExpress
    • LLDP (Link Layer Discovery Protocol) support
    • Optional enhanced host authentication through ethernet switches
    • Faster boot
    • Debug shell on tty2
    • UEFI and legacy x86 boot modes supported
    • nbfs remains available
  • IPMI now honors site.syspowerinterval and site.syspowermaxnodes to optionally throttle rpower operations.
  • IPMI allows for new attribute site.ipmidispatch to be set to '0' to disable dispatching the IPMI hardware operations to the service nodes.
  • Allow for per-node override of tftpdir to allow for mount-managed service node tftp content or surrogate service node tftp content
  • KVM plugin now allows the administrator to specify a more traditional password for VNC/SPICE access using vm.vidpassword
  • Improved logging of node discovery attempts
  • General support of new IBM 'M4' generation hardware
  • Enhanced out-of-band inventory data on new IBM system x servers
  • Can now rpower suspend select IBM servers
  • OS installers are now allowed to go graphical if no text console is specified
  • ESXi5 support now includes stateful install in addition to the previous stateless boot support
  • xnba now supports UEFI boot of relevant operating systems (ESXi 5, SLES11, RHEL6, Win2k8, Win7)
  • Aids to help monitor Mellanox IB networks. See [Managing_the_Mellanox_Infiniband_Network].

Linux

  • The servicenode postscript now calls xcatserver and xcatclient for Linux. You do not need all three in the postscript list. (It will not hurt if they are there.) The install or update of xCAT will cleanup the list in the postscripts table. So where you previously saw "servicenode,xcatserver,xcatclient" you will only see "servicenode".
  • Switch to using the version of tftp that comes in the distro, instead of the atftp-xcat RPM in the xcat-dep tarball.
  • Load distro DVD disc id's from a separate file (/opt/xcat/lib/perl/xCAT/data/discinfo.pm), instead of them being in the plugins. This makes it easier for users to modify this file to try out a new version of a distro.
  • Removed the requirement to setup and use FTP to download the postscripts to the nodes. Uses httpd now.
  • Documentation on how to setup RAID1 for Linux diskful installation. See [Use_RAID1_In_xCAT_Cluster]
  • General support for RHEL 6.2
  • xCATs installation/configuration IBM HPC products on RHEL 6.2 is experimental until 2.7.1
  • Rolling updates on SLES 11 SP1 and RHEL 6.2 is experimental until 2.7.1
  • Mellanox IB QDR support on IBM system p and system x servers
  • Support for Nagios monitoring plugin

AIX

  • mkdsklsnode supports new option ( -d) to Only define the NIM resources on the service nodes.
  • Support for using NFS v4 with AIX diskless nodes.
  • Experimental code for real-time service node fail over on AIX. This code depends on some fixes in AIX that are not available yet, so this code is just for investigation on a non-production system.
  • dumpxCATdb - added -b option for those running the DB2 database. It will use the DB2 database dump utilities to create a binary backup of the entire DB2 xCAT instance. See also the xCAT DB2 doc. Also supported on p775 linux.
  • restorexCATdb - added -b option for those running the DB2 database. This will restore the database from the binary backup taken using DB2 Utilities. Also supported on p775 linux.

Key Bugs Fixed

See the xCAT 2.7 SourceForge bugs.

Restrictions and Known Problems

2.6.11 Release Notes


2.6.10 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

  • Support for mkdsklsnode -n in AIX clusters to more quickly switch nodes to a new image
  • Huge page support in p775 clusters
  • Add site table syspowerinterval attribute for booting p775 CECs.
  • More performance improvements for DFM in p775 clusters
  • Redundant service network support for p775 clusters
    • Requires the HPC hardware server version 1.1.0.2 or higher for AIX and 1.1.0.1 or higher for Linux
    • On AIX, this feature has been FVT'd (Functional Verification Test), but has not yet been System Tested.
    • On Linux, this feature has been FVT'd and scale tested.
  • snmove can now copy statelite persistent files to the backup service node
  • RHEL 6.1 support for non-p775 clusters
  • RHEL 6.1 support for p775 clusters(experimental)
  • RHEL 6.2 early support on system p
  • Added binary backup option to dumpxCATdb for DB2 to do database backups more efficient for large databases.

Key Bugs Fixed

For additional bugs fixed, see Tracker Bugs

Restrictions and Known Problems

  • For AIX diskless nodes, if you define /var/adm/ras/errlog as a statelite persistent file in the litefile table, default entries in crontab like the following may cause significant slow downs on the service nodes for a few minutes when those entries run (due to high disk usage on the SNs):

    0 11 * * /usr/bin/errclear -d S,O 30
    0 12 *
    * /usr/bin/errclear -d H 90

    You may want to reduce the frequency that these commands are run.

  • When powering on multiple p775 CECs or powering on LPARs of multiple p775 CECs via rpower, set site.powerinterval to 30 before running rpower. See https://sourceforge.net/tracker/?func=detail&aid=3426911&group_id=208749&atid=1006945 for details.

  • The rnetboot command for p775 LPARs won't correctly set the default bootlist in AIX diskless nodes so the LPAR can be rpower'd next time. Also, the default bootlist can sometimes change after the user sets it. The workaround for now is to always run rbootseq before running rpower on for p775 LPARs. See https://sourceforge.net/tracker/index.php?func=detail&aid=3435969&group_id=208749&atid=1006945 for details.
  • Lose DB2 root setup on Linux when updating xCAT rpms on MS and SN's. See defect 3474407 for manual fix. https://sourceforge.net/tracker/?func=detail&aid=3474407&group_id=208749&atid=1006945
  • SF defect 3477804. This defect affects you, if you are running DB2 and LoadLeveler using DB2 on the service nodes. If you run updatenode <servicenodes> -S on Linux or AIX, and you find that the /etc/odbc.ini and odbcinst.ini have been reinitialized back to defaults, then you will need to xdsh <servicenodes> /opt/xcat/bin db2sqlsetup -C -o, to set these files back up for LoadLeveler to be able to access the database.

2.6.9 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

  • Increased timeout value for query of LEDs from Bladecenter
  • AIX 7.1.1 support
  • Ability to set BSR setting on p775
  • Performance improvements for DFM p775 hardware control
  • Stateful deployment of AIX over HFI
  • Postscript to set up disk mirroring on an AIX stateful node
  • Mixed clusters of system x and system p hardware running SLES: [Mixed_Cluster_Support_for_SLES]
  • xcatsnap improvements
  • Hardware replacement procedure documentation(P775A+ part)

Key Bugs Fixed

  • 3412934 - Improve output from genimage
  • 3414742 - xdsh -K did not work if userid was in LDAP (not in /etc/passwd)
  • 3430879 - improve configeth sample postscript and documentation
  • 3411554 - CentOS 6 image gen is missing dracut link
  • 3377379 - genimage plugin hangs when postinstall script prompts
  • 3412259 - lsslp couldn't get parent for some cecs
  • 3419608 - mkdsklsnode fails first time for hfi_net
  • 3427796 - additional options needed in aixvgsetup
  • 3390380 - sles10SP4: genimage error with dbus not found
  • 3428290 - xcat fork memory errors with rbootseq
  • 3428344 - Confighfi support for Torrent 2.1 HFI netwk
  • 3398468 - AIX diskless CNs miss paging space in large p7
  • 3426911 - support for site.powerinterval added

For additional bugs fixed, see Tracker Bugs

Restrictions and Known Problems

  • The rpm package conserver-xcat was added in the dependency package to replace conserver. If encountering the issue that conserver-xcat conflicts with conserver when updating of xCAT from 2.6.9 to 2.7, remove the conserver manually first and try again.
  • When powering on multiple p775 CECs or powering on LPARs of multiple p775 CECs via rpower, set site.powerinterval to 30 before running rpower. See https://sourceforge.net/tracker/?func=detail&aid=3426911&group_id=208749&atid=1006945 for details.
  • The rnetboot command for p775 LPARs won't correctly set the default bootlist in AIX diskless nodes so the LPAR can be rpower'd next time. Also, the default bootlist can sometimes change after the user sets it. The workaround for now is to always run rbootseq before running rpower on for p775 LPARs. See https://sourceforge.net/tracker/index.php?func=detail&aid=3435969&group_id=208749&atid=1006945 for details.
  • AIX is shipping it's own chdef and psh command in AIX 7.1.1. With normal installation they fall first in the path, so you may be surprised by the fact your man page and command are not what you expect when you run chdef of psh. You are going to need to update your profile to put the xcat paths before the OS chdef and psh, and also modify the MANPATH to put the xCAT man page first. https://sourceforge.net/tracker/?func=detail&aid=3424615&group_id=208749&atid=1006945

2.6.8 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

These are the changes since the xCAT 2.6.7 release.

New Function and Changes in Behavior

  • New Commands. See corresponding man pages for details:
  • Enhanced Commands. See corresponding man pages for details:
  • Fully support parameters for postscripts
    • Parameters can be added in the postscripts table like script1,script2 p1 p2,script3...
    • Parameters can be included in the updatenode -P command. For example: updatenode noderange -P "script p1 p2"
  • HA EMS on Power 775
  • External NFS server support on Power 775
  • Add /proc file system by default for AIX diskless nodes
  • xcatsetup supports new syntax like f[1-6]c[01-12]p[01,05,09,13,17,21,25,29] inhostname-range keyword
  • xcatsetup to support BPA name using Frame nd CEC
  • SLES 10 SP4 diskful installation support
  • xcatdebug, See man page for more details and [Debugging_xCAT_Problems]
  • Disk mirroring support with SLES 11 [Use_RAID1_In_xCAT_Cluster]

Defect fixes

  • DNS forwarder cannot work on AIX platform
  • rspconfig query for deconfigured resources on Power 775

Restrictions and Known Problems

  • For AIX diskless nodes, the xcatmaster attribute in the database must be an ip address of the Service Node as known by the node. The current NIM creation of a diskless image builds the /etc/hosts file with only the long hostname of the service node, so putting a short hostname in the xcatmaster attribute will not resolve on the node. If there is no hierarchy, then the site table master attribute is used and that also must be the ip address of the Management Node as known by the nodes. This affects the setup of syslog and node to node passwordless ssh support on the AIX diskless nodes during install.

2.6.6 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

These are the changes since the xCAT 2.6.2 release. Note: for reasons too difficult to explain here, the xCAT version number went straight from 2.6.2 to 2.6.6. There were no releases named 2.6.3, 2.6.4, or 2.6.5.

New Function and Changes in Behavior

  • New Commands. See corresponding man pages for details:
    • swapnodes - for Power 775 Availability Plus support
  • Enhanced Commands. See corresponding man pages for details:
    • lsdef
    • "-c" flag to display node information in compressed mode
    • "--osimage" flag to list the osimage information for the node
    • dumpxCATdb and restorexCATdb will honor a new site.skiptables attribute to limit the tables that are dumped and restored
    • updatenode added support to just sync files to the service nodes ( -f flag)
  • Support for Power 775 Clusters:
    • Direct FSP Management(DFM) initial release. See [XCAT_Power_775_Hardware_Management#xCAT_Direct_FSP_and_BPA_Management_Capabilities] for more details on its capabilities, and see [XCAT_Power_775_Hardware_Management#Downloading_and_Installing_DFM] for information about installing it.
    • rpower - Transition low power states, exit/enter rack standby, on/off/query of power state for CEC and LPAR
    • rcons - Remote Console
    • rflash - Firmware support for FSP and BPA
    • rinv - Get the firmware level of FSP and BPA; get the deconfigured resource of CEC
    • rvitals - Display LCD values; Get the rack environmental information
    • getmacs - HFI MAC Address information collection
    • mkhwconn/rmhwconn - Make and remove FSP and BPA hardware connections
    • lshwconn - List hardware connection status
    • rnetboot - Remote network boot
    • lsvm/chvm - LPAR list, creation and removal; I/O slot assignment, get and set of LPAR name
    • rbootseq - Sets the net or hfi device as the first boot device for the specified PPC LPARs
    • rspconfig - FSP and BPA password support; get and modify the frame number, get and set of frame name
    • Redundancy FSP/BPA support
    • Energy management for Power 775 servers. FFO (Fixed Frequency Override) for Power 775 servers that the accurate CPU frequency can be set to the server. See the renergy command.
    • Boot over HFI for AIX 7.1 and RHEL6
    • Power 775 HFI mac address failover support
    • Hardware replacement procedure for Power 775 Availability Plus. See [Power_775_Cluster_Recovery]
  • DB2 WSE support. See [Setting_Up_DB2_as_the_xCAT_DB]
  • iSCSI dump on AIX 7.1, kdump on RHEL6
  • HA MN support(experimental). See [Shared_Disks_HA_Mgmt_Node]
  • External NFS server support (experimental). See [External_NFS_Server_Support_With_Linux_Statelite]
  • Service node manual failover (experimental)
  • Locating all of the node deployment related files using nodetype.provmethod=osimage on Linux. So far, this has only been fully tested for stateless nodes. Statelite is supported per the instructions in the Statelite documentation, [XCAT_Linux_Statelite]. We will test it for full disk install and more statelite testing in the future. See [Using_Provmethod=osimagename]
  • HPC integration enhancements:
  • Rolling Update support for pLinux clusters and AIX clusters. See [Rolling_Update_Support]
  • ddns plugin is the default dns handler. See [Cluster_Name_Resolution]
    • The optional ddns plugin available for dynamic DNS support from xCAT 2.5 is not optional anymore. It is the only one shipped, and supported as of 2.6. The bind.pm plugin has been removed. If you want to keep the existing DNS settings made by xCAT BIND, then you should not run any "makedns" commands. If you want to use the Dynamic DNS feature then you must run "makedns -n" to refresh the DNS settings.
    • makedns needs the /etc/resolv.conf on management node to have the mn's IP address specified as nameserver (from site.master) and cluster domain ( from site.domain) as a search path. If the compute nodes also needs to have name resolution to hosts outside the cluster, add the external nameservers addresses to the site table forwarders attribute.
    • A warning will be issued when running makedns, if the nameserver or search paths are not set in /etc/resolv.conf. For AIX, this message is in error and should be checking for domain site.domain clause, and not search site.domain. It is just a warning and processing will continue.
    • xCAT sets site.dnshandler to ddns automatically for two scenarios:
    • Fresh install for xCAT 2.6
    • Update install for xCAT 2.6 from an existing lower level xCAT version.
    • If you restored xCATdb from a earlier backup(xCAT 2.5.x or earlier) after xCAT 2.6 is installed, it would overwrite or remove site.dnshandler, you need to manually set site.dnshandler=ddns after the restore, otherwise, makedns can not work.
  • DNS hierarchical support: DNS on the service nodes forward unknown DNS request to management node
  • Automatic creation of resolv.conf files on AIX nodes when using DNS.
  • Use /etc/hosts as the name resolution when site.nameservers is blank
  • New node attribute 'hwtype'. See node definition
  • Disk mirroring setup in xCAT Linux cluster. [Use_RAID1_In_xCAT_Cluster]
  • xCAT web interface initial release(experimental).
  • Soft mount option support for statelite persistent files/directories.
  • SLES 11 SDK iso image support
  • Support pattern in genimage and updatenode
  • genimage generic command (/opt/xcat/bin/genimage) now runs under the xcatd daemon as a plugin. You may notice changes in the interface and output displayed. If there is any prompting, for example acceptance of licenses, during your genimage; you can no longer run the generic genimage command, and must run the specific genimage script from the /opt/xcat/share/xcat/netboot/<os>/genimage directory.
  • genimage specific commands ( /opt/xcat/share/xcat/netboot/<os>) creates the ifcfg-* file only if the -i flag is used, and has been changed to create with ONBOOT=no
  • Support for preserving ODM data on diskless-stateless AIX nodes.
  • Support for the use of dhcp for AIX installs. (Includes makedhcp enhancements.)
  • A new sample postscript, ("make_sn_fs"), that may be used to create and mount local filesystems on the xCAT service nodes when they are installed.

Restrictions and Known Problems

  • genimage generic command (/opt/xcat/bin/genimage) now runs under the xcatd daemon as a plugin. If there is any prompting, for example acceptance of licenses, during your genimage; you can no longer run the generic genimage command, and must run the specific genimage script from the /opt/xcat/share/xcat/netboot/<os>/genimage directory.
  • DNS fowarders can not work on AIX - bug 3391271

    The workaround is to dig the forwarder's dns records into /var/named/db.cache file and restart named service.
    For example:
    If your forwarder is 9.114.1.1, then run commands below on your xcat management node:
    dig @9.114.1.1 . ns >>/var/named/db.cache
    stopsrc -s named
    startsrc -s named

  • Power 775 support - On AIX after reboot of the EMS, the ISNM software will not automatically start and LoadLeveler will fail to start. In fact any 32 bit application that uses the DB2 database will have problems connecting to the database. This is a DB2 APAR, it is referenced in the defect and the problem will be fixed when V9.7.5 fix pack is available and applied.

    See the following for instructions to restart the HPC stack:
    [Setting_Up_DB2_as_the_xCAT_DB#Power_775_Special_instructions_for_Reboot_of_EMS_on_AIX]
    See the following defect for details: bug 3391193

2.6 Release Notes


This is the summary of what's new in this release. Or you can go straight to [Download_xCAT].

New Function and Changes in Behavior

  • New Commands. See corresponding man pages for details:
    • lsxcatd - List daemon information
    • xcsv - Reformat output of rvitals and similar in CSV format
    • imgcapture - capture image from running node to create a stateless/statelite image - only for linux
  • Enhanced Commands. See corresponding man pages for details:
    • tabdump -w - allows selective dump of table rows
    • xdcp - now supports the automatic running of postscripts after files are rsync'd to the nodes. See the xdcp man page and Sync-ing_Config_Files_to_Nodes#postscript_support.
    • makeroutes - added ability to specify routes for compute nodes
    • snmove - Service node manual failover - added support for AIX and enhancements for linux
    • monshow - Now also shows RMC events (in addition to RMC performance info)
  • Support for statelite semantics on top of a ramdisk diskless node
  • Support for DB2 on RHEL 6
  • xCAT now supports the automatic setup of DB2 client on the Service Node during Service Node install.
  • Documentation of how to manually fail over an xCAT management node to a backup using a shared disk
  • Support on AIX for using a separate NFS server for diskless nodes
  • Documented [External_NFS_Server_Support_With_Linux_Statelite]
  • HPC integration set up of login nodes - pkg lists, postscripts, etc. for what needs to be installed on/configured for a login node.
  • Support for alt_disk install and/or multibos on diskfull AIX nodes
  • Exploitation of batched event hierarchical support (attributes on the Condition class that indicate when a batched event file is ready to be processed)
  • Documented [Hints_and_Tips_for_Large_Scale_Clusters]
  • Support for setting up kdump for linux diskless nodes
  • OS dump via iSCSI AIX stateless and statelite nodes. The iscsi dump support is described in section 3 of the "xCAT AIX Diskless Nodes" doc.
  • AIX 71B/61N support for p7 & p6
  • Added documentation for IB support on system p on RHEL 6
  • Ability to set up the compute node's service node as the default gateway by setting networks.gateway] to "<xcatmaster>"
  • A new attribute, site.excludenodes, to specify a list of nodes/groups that should always be excluded from all xcat commands (e.g. the list of nodes the currently have hardware problems).
  • Enhanced IPv6 support
    • ddns plugin will now push IPv6 /etc/hosts entries into DNS
    • DHCPv6 bindings based on client DUID-UUID for ISC DHCP 4.x (used in RHEL6)
    • Windows, ESXi, and RHEL6 support for using DUID-UUID on the client side
  • Per network dynamic dns suffix allows each network to potentially have a different IPv6 suffix. Useful for dual stack and multihomed nodes in dynamic address use (e.g. node1.cluster.example.com, node1.ib.cluster.example.com, node1.ipv6.cluster.example.com is now possible).
  • Enhancements to xcatmon/appstatus to monitor the HPC application (GPFS, LAPI and LoadLeveler) status in xCAT cluster. Added some new sample scripts for lcmd and dcmd. See Monitoring HPC application status.
  • Partial support for vSphere 5

Limitations and Known Issues

For additional issues, see Tracker Bugs


Related

Wiki: Main_Page
Wiki: XCAT_2.8.4_Release_Notes

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks