|
From: Walker, B. J <bru...@hp...> - 2004-12-28 18:38:47
|
En Chiang, There is no mention of DRBD in release notes below. Could you please clarify to the community where we stand and when we will have a package (with instructions) available. Thanks, bruce > -----Original Message----- > From: ssi...@li...=20 > [mailto:ssi...@li...] On=20 > Behalf Of Lee, En Chiang (STSD) > Sent: Wednesday, December 22, 2004 3:12 AM > To: ssi...@li... > Cc: ssi...@li...;=20 > ssi...@li... > Subject: [SSI-devel] OpenSSI 1.2.0 released for Fedora Core 2 >=20 >=20 > The OpenSSI 1.2.0 stable release for Fedora Core 2 is now=20 > available from > OpenSSI.org. The versions for Debian and Red Hat 9 will follow in > several days. >=20 > The release notes are below. >=20 > Enjoy! >=20 > En Chiang >=20 > --- > OpenSSI 1.2.0 is a stable release, suitable for production use. These > release notes are compilation of the notes for 1.1.0 and=20 > 1.1.1, for the > convenience of users who have not upgraded since 1.0.0. A=20 > description of > what's changed since 1.1.1 can be found at the bottom. >=20 > This release has versions for Fedora Core 2 ("FC2"), Debian=20 > testing, and > Red Hat 9 ("RH9"). The OpenSSI 1.2.x series will be the last set of > releases for RH9, so you should only use RH9 if you're upgrading from > 1.0.0 or 1.1.x for RH9, and you're unable to do a fresh install on FC2 > or Debian. >=20 > The OpenSSI kernel is now based on the most recent Fedora=20 > Core 1 ("FC1") > kernel (2.4.22-1.2199.nptl). This is true for all distributions, since > it is difficult to maintain multiple versions of the OpenSSI=20 > kernel. It > might seem strange that OpenSSI 1.2.0 runs on FC2, but it is based on > the FC1 kernel. This is because the FC1 kernel is based on the Linux > 2.4 kernel, whereas FC2 is based on 2.6. There is currently a=20 > project to > port OpenSSI to the 2.6 kernel, but it will not be ready for=20 > a few more > months. When it is ready, it will be part of OpenSSI 1.9/2.0. >=20 > There are several features that have been developed for OpenSSI 1.2 > since the last stable release: 1.0.0. One of them is performance > enhancements for the Cluster File System ("CFS"). Now it not=20 > only caches > remote reads, but it also caches remote writes, while still=20 > maintaining > a coherent view of the filesystem across the cluster. Furthermore, CFS > now does asynchronous remote read-aheads of data blocks that programs > might want, so that they are already cached locally by time=20 > the program > wants them. Hopefully you will notice a significant performance > improvement in your filesystem intensive application! >=20 > Another new feature is atomic migration of a group of processes. The > group could either be a POSIX process group or the "threads" of a > multi-threaded application (on Linux, each thread is a full=20 > process). To > migrate a POSIX process group, call the migrate command with the > negative PID of the process group leader. These semantics are very > similar to signaling a process group. To migrate a thread group, call > the migrate command with the PID of any "thread" in that group. With > both forms of group migration, either the entire group migrates or it > does not. If any process in the group is unable to migrate for any > reason, the entire group will remain on the old node. >=20 > Another process migration enhancement is the ability to migrate a > process while it is holding file record locks. These locks=20 > will continue > to be held during the migration and after the process=20 > continues running > on the new node. >=20 > LVS-NAT can now be used with OpenSSI. Linux Virtual Server=20 > ("LVS") is a > third-party open-source project that load balances TCP=20 > connections among > the nodes in a cluster. For a long time, LVS has been integrated with > OpenSSI, making it easier to manage than an LVS cluster=20 > without OpenSSI, > but only the Direct Routing ("DR") feature of LVS was supported. DR > allows load balancing in a cluster where every node has a direct > connection to the external network (as well as a network=20 > connection to a > private switch for the cluster interconnect, as recommended in the > OpenSSI installation instructions). Unfortunately, a security=20 > feature in > the Linux kernel prevents DR from being used in a cluster where only > some nodes are connected to the external network. For these=20 > situations, > the Network Address Translation ("NAT") feature of LVS should be used, > and it is now supported on OpenSSI. >=20 > Note that LVS-NAT is different from the NAT that you would use for > making outbound connections from a private IP address. LVS-NAT is for > load balancing inbound connections from a public IP address among a > cluster of machines that are all connected to a private=20 > network, such as > an OpenSSI cluster interconnect. Of course, LVS-NAT requires that > potential director nodes are connected to both the external=20 > network and > the cluster interconnect, so that traffic can move between the two > networks. >=20 > The 'fast' and 'fastnode' commands have been added. 'fastnode' returns > the node number for the least-loaded node in the cluster, as=20 > determined > by the process load-leveling algorithm. 'fast' executes a=20 > command on the > least-loaded node. Read the man pages for these commands for more > information. >=20 > Several files were added to /proc/cluster/: nm_rate, nm_log_threshold, > and nm_nodedown_disabled. nm_rate can be used to alter how often node > monitoring messages are exchanged (default is 1 per second)=20 > and how long > before a node is declared down (default 10 seconds). nm_log_threshold > indicates how many monitoring messages can be missed before a kernel > warning is generated (default 2). nm_nodedown_disabled can be set to > disable nodedown detection, which is useful if you need to enter the > kernel debugger on one of the nodes. Previously, you had to=20 > recompile an > OpenSSI kernel to change one these values. Now you can do it by simply > writing a new value into one of these /proc/cluster files. >=20 > The top command was enhanced for this release by Roopa Prabhu. By > default, it adds an execution node number column and only displays > clusterwide information in the header. The node number column replaces > the mem % column, which is potentially confusing. When top is run in > localview mode (i.e., `localview top'), it limits the list of=20 > processes > to just those that are running on the local node, and it displays all > the same information as the base version of top. >=20 > To improve performance, the init failover state file was moved from > /etc/initstate to /cluster/init/initstate. This avoids the need to > constantly hit the large /etc directory. >=20 > Since the last development release (1.1.1), OpenSSI now has=20 > the ability > to automatically move loadleveled process off a node that is=20 > gracefully > shutting down due to a clusternode_shutdown call. >=20 > Another new feature is an interactive 'e' command for top,=20 > which prompts > the user for a node number and displays only processes on that node. > This new feature is only available when top is run in defaultview (not > localview). >=20 > A README was added for configuring clusterwide NFS client mounts. >=20 > Several interface changes to HA-LVS include a new > /proc/cluster/lvs_internal_gw file, as well as changes to make > /proc/cluster/lvs_routing use the seq_file interface. The files > /usr/sbin/clusterip.sh and /etc/default/lvs_routing no longer exist. >=20 > There have been many bug fixes. See the ChangeLog for more details. >=20 >=20 >=20 >=20 > ------------------------------------------------------- > SF email is sponsored by - The IT Product Guide > Read honest & candid reviews on hundreds of IT Products from=20 > real users. > Discover which products truly live up to the hype. Start reading now.=20 > http://productguide.itmanagersjournal.com/ > _______________________________________________ > ssic-linux-devel mailing list > ssi...@li... > https://lists.sourceforge.net/lists/listinfo/ssic-linux-devel >=20 |
