ci-linux-devel — Discussion about Cluster Infrastructure for Linux

[CI] Kernel Panic

[Eduardo S. M. <ed...@ci...>]

Hello,

My system is a cluster pack of DL380 with MSA500 G2. DL380 is two
processors with Hyperthreading. 

I installed CI software and I boot with CI kernel. I got a kernel panic
just before run cluster_start.

I got this message:

Kernel panic: ics_llunack_open: bind error -99
Instruction(i) breakpoint #0 at 0xc0122760 (adjusted)
0xc0122760 panic_hook:	int3
Entering kdb (current=0xf7062000, pid 625) on processor 0 due to
Breakpoint @ 0xc0122760

Any idea about this problem?

Best Regards,

Eduardo Sanz

[CI] [2.6] CI port to UML

[Krishnakumar. R <rk...@in...>]

Hi,

CI for 2.6 has been ported to UML.

A HOWTO on running virtual CI cluster using
UML is available at the location:
http://ci-linux.sourceforge.net/contrib/uml/docs/ci-uml.html

Also available are the patches and binaries.
More information about them can be found 
within the HOWTO.

Thanks and Regards,
KK.

[CI] [ ci-linux-Feature Requests-1058615 ] functionality to reset ics statistics

[SourceForge.net <no...@so...>]

Feature Requests item #1058615, was opened at 2004-11-02 11:19
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405833&aid=1058615&group_id=32543

Category: None
Group: None
Status: Open
Priority: 5
Submitted By: Aneesh Kumar K.V  (kvaneesh)
Assigned to: Nobody/Anonymous (nobody)
Summary: functionality to reset ics statistics

Initial Comment:
To compare the ICS statistics running tests under
various conditions, it is nice to have a way to reset
the counters to zero before you start the tests.






----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405833&aid=1058615&group_id=32543

Re: [CI] ci INSTALL,1.4,1.5

[Brian J. W. <Bri...@hp...>]

Aneesh Kumar K.V wrote:
> It should be INTERFACES=eth0
> 
> and i guess we dropped ICS_ROUTE . We are now configuring everying using 
> ifconfig from user space IIUC.

Fixed.

Thanks,

Brian

Re: [CI] ci INSTALL,1.4,1.5

[Aneesh K. K.V <ane...@hp...>]

Brian J. Watson wrote:
> Update of /cvsroot/ci-linux/ci
> In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv27803
> 
> Added Files:
> 	INSTALL 
> Log Message:
> Make CI instructions generic for all distros
> 
> 
> --- NEW FILE: INSTALL ---
> Installing Cluster Infrastructure ("CI")
> 
> These instructions describe how to install a CI cluster with 
> minimal hardware requirements. All you need is two or more 
> computers, connected with a private ethernet network. This 
> network is called the "interconnect", and should be private for 
> security and performance reasons.  Each individual computer in 
> the cluster is called a "node".
> 
> CI contains a subset of the functionality of OpenSSI 
> (http://OpenSSI.org). It provides an in-kernel CLuster Membership 
> Service ("CLMS") and Internode Communication Subsystem ("ICS").
> 
> CLMS keeps track of nodes joining and leaving the cluster and 
> guarantees an identical view of cluster membership on every node.
> An included library, libcluster, allows access to CLMS information 
> from user-mode. Various man pages document the libcluster API.
> There are also several programs that provide command-line access
> to the libcluster API.
> 
> CLMS depends on ICS, which is a protocol for passing Remote 
> Procedure Calls ("RPCs") between nodes, using node numbers as 
> addresses. ICS stacks on top of TCP, and is architected so that it 
> can be ported on top of other reliable transport protocols. There 
> is currently no API for using ICS from user-mode, although 
> docs/enhancing.txt discusses how to use both ICS and CLMS from 
> kernel-mode.
> 
> These instructions assume you are doing a fresh install of 
> CI. They must be followed for each node you wish to add to 
> the cluster.
> 
> 1. Install your Linux distribution.
> 
> 2. When configuring your firewall, do one of the following:
> 
>   (a) designate as "trusted" the interface for the cluster interconnect
> 
>   (b) open the port for tftp-server (port 69)
> 
>   (c) disable the firewall
> 
> 3. Extract the CI tarball and run the ./install script. It will
>   install the necessary packages, and prompt you to reboot.
>   Select the CI kernel during boot.
> 
> 4. Create a /etc/cluster.conf file. This will contain the boot 
>   parameters used by cluster_config to add the node to the cluster. 
>   Each boot parameter should be on a line of its own.
> 
> 	IFCONFIG=eth<x>:<ipaddr>:<netmask> 
> 	CLUSTER_MASTER=<y>:<clms_ipaddr1>[,<z>:<clms_ipaddr2>]
> 	CLUSTER_NODENUM=<a>
> 	[ICS_ROUTE="default"|<host ipaddr>:<gateway ipaddr>:<metric>]
> 
>  

It should be INTERFACES=eth0

and i guess we dropped ICS_ROUTE . We are now configuring everying using 
ifconfig from user space IIUC.

-aneesh

[CI] CI with Qemu.

[Aneesh K. <ane...@gm...>]

The below snapshot tells it all 

http://ci-linux.sourceforge.net/contrib/qemu-ci.png

This is a two node CI linux cluster running with qemu. 
( http://fabrice.bellard.free.fr/qemu/ ) 

It is running debian testing  and 2.6.8.1 kernel  patched with CI.
The below document was really helpful in getting everything running. 

http://www.geocities.com/paul_phoenix2779/openmosix-dojo/Qemu-openMosix.doc.txt

-aneesh

[CI] netdump patches for 2.6.

[Aneesh K. <ane...@gm...>]

Hi,

I have placed the netdump patches for 2.6 at 
http://ci-linux.sourceforge.net/contrib/linux-netdump.diff

-aneesh

[CI] Two node CI/2.6 cluster

[Brian J. W. <Bri...@hp...>]

Hi Aneesh,

Good news!! I found the deadlock that was preventing the second node 
from joining. Due to a locking change in flush_signals(), the icssvr 
daemon was attempting to grab the same spinlock twice with interrupts 
disabled. I removed the extraneous spinlock code from the CI wrapper 
function, and a two-node CI cluster now works! I even did a couple of 
CLMS master failovers.

The modified patch is available at
http://ci-linux.sourceforge.net/contrib/ci-2.6.6.ak.aug11.patch.bz2

I still need to review the changes you made to get CI/2.6 to build. I'll 
be on vacation until next Tuesday, so I'll review them after I get back. 
Then I'll create a new branch in the CI repository for this stuff.

Best regards,

Brian

[CI] CI on 2.6

[Brian J. W. <Bri...@hp...>]

Hi Aneesh,

I built and booted your latest CI/2.6 kernel on Fedora Core 2 for i686. 
One node comes up fine. When I attempted to join the second node, both 
nodes hung. IIRC, you haven't tested yet with two nodes.

Have you tried building CI/2.6 with kdb, yet? That would be a useful 
tool to have.

Regards,

Brian

Re: [CI] New NIC

[Brian J. W. <Bri...@hp...>]

Marc-Andre Heroux wrote:
> Dear Openssi developers,
> 
> I am running a cluster node1 on a RedHat 9 machine.  
> 
> I would like to add a new network card module for a new node, node2, the
> RTL8139 NIC.
> 
> Is there an easy way to proceed?

Add the line 'alias eth-extra rtl8139' to /etc/modules.conf (assuming 
rtl8139 is the name of the driver). Run the following command to rebuild 
your ramdisk:

mkinitrd -f --cfs /boot/initrd-<kernel_version>.img <kernel_version>

Run ssi-ksync to distribute the updated ramdisk.

You can determine what kernel versions you have installed by looking in 
/lib/modules. The <kernel_version> for the OpenSSI 1.0.0-rc6 release is 
2.4.20-31.9_ssi_5smp.

Hope this helps,

Brian

[CI] New NIC

[Marc-Andre H. <mah...@tc...>]

Dear Openssi developers,

I am running a cluster node1 on a RedHat 9 machine.  

I would like to add a new network card module for a new node, node2, the
RTL8139 NIC.

Is there an easy way to proceed?

Best Regards to all of you,

Marc-Andre from Montreal, Qc.

[CI] [ANNOUNCE] openais project to impelement SA Forum AIS

[Steven D. <sd...@mv...>]

This is slightly off-topic, but for those interested in high
availability clusters might find this project helpful:

Project Announcement: Open AIS (http://developer.osdl.org/dev/openais)

It is with great pleasure that I announce the openais project.  The
openais project implements a production quality BSD licensed
implementation of the SA Forum's Application Interface Specificaton for
high availability compute clusters.  The target operating system is
Linux.

The openais project is the first published AIS implementation to provide
strong membership guarantees in the AIS cluster membership API.  Another
first is the first published implementation anywhere of the AIS
availability management framework API.  Finally the implementation
includes the first published AIS checkpointing API to achieve ~70%
network-wire-speed with one node checkpoint writing eight replicas.

The openais project is the first published AIS implementation to use
cutting edge research in the virtual synchrony group messaging model. 
Virtual synchrony provides several benefits such as using standard
UDP/IP, utilization of available hardware multicast, and avoiding
two-phase commit protocols.  The major advantage of this model is
performance: the checkpoint service can checkpoint 8.8 mb/sec of data on
a 100 mbit network to two replicas and 6.3 mb/sec of data to twelve
replicas.  Another significant advantage of this technology is 100%
correct operation during faults, partitions, and merges.

Please join us in evaluating and contributing to the openais project. 
With your help, we can join together to develop a world-class
implementation of SA Forum's Application Interface Specification.

The quickstart documentation can be found:

http://developer.osdl.org/dev/openais/src/QUICKSTART

The main project page can be found:

http://developer.osdl.org/dev/openais

To subscribe to the mailing list, visit:

http://lists.osdl.org/mailman/listinfo/openais

[CI] [ ci-linux-Bugs-532588 ] Tryied it on an Alpha EV67 !??! -Part 2

[SourceForge.net <no...@so...>]

Bugs item #532588, was opened at 2002-03-20 09:23
Message generated for change (Settings changed) made by bjbrew
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405830&aid=532588&group_id=32543

Category: None
Group: None
>Status: Closed
Resolution: None
Priority: 5
Submitted By: Oliver Pitzeier (opitzeier)
Assigned to: Nobody/Anonymous (nobody)
Summary: Tryied it on an Alpha EV67 !??! -Part 2

Initial Comment:
I also get this with the cluster-tools:
[root@clusternode1 cluster-tools-0.5.6]# make install
for i in libcluster cmd ; do cd /root/cluster-tools-
0.5.6/$i ; make install ; done
make[1]: Entering directory `/root/cluster-tools-
0.5.6/libcluster'
gcc -I../include -Wall -c -o ssisys.o ssisys.c
ssisys.c: In function `ssisys':
ssisys.c:24: inconsistent operand constraints in an 
`asm'
make[1]: *** [ssisys.o] Error 1
make[1]: Leaving directory `/root/cluster-tools-
0.5.6/libcluster'
make[1]: Entering directory `/root/cluster-tools-
0.5.6/cmd'
gcc -I../include -Wall -O2 -D_GNU_SOURCE -c -o 
ssisys.o ssisys.c
ssisys.c: In function `ssisys':
ssisys.c:24: inconsistent operand constraints in an 
`asm'
make[1]: *** [ssisys.o] Error 1
make[1]: Leaving directory `/root/cluster-tools-
0.5.6/cmd'
make: *** [install] Error 2


-Oliver

----------------------------------------------------------------------

Comment By: Oliver Pitzeier (opitzeier)
Date: 2002-03-22 00:44

Message:
Logged In: YES 
user_id=156662

Well, hi again Aneesh!
This worked as well!
But see the other bug-note... I'm still having problems...

:o)

-Oliver

----------------------------------------------------------------------

Comment By: Aneesh Kumar K.V  (kvaneesh)
Date: 2002-03-20 21:03

Message:
Logged In: YES 
user_id=230991

I guess this is a know problem which i have added in  all my
previous mail. It  is the build issue that i pointed before
with cluster tools. You need to replace the unistd.h file in
the cluster tools with the  unistd.h file from the alpha
directory. 

 BTW try the latest clustertools. 

cd cluster-tools-0.6.0

cp /usr/src/linux/include/asm-alpha/unistd.h
include/asm/unistd.h 

( I guess we should  all these header files to pick from the
linux kernel ) 


-aneesh 



----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405830&aid=532588&group_id=32543

[CI] The trunk for the CI and OpenSSI repositories

[Brian J. W. <Bri...@hp...>]

A file should only be added to the trunk for the CI and OpenSSI 
repositories if it's distro-neutral (i.e., it's absolutely identical 
between the Debian and RH branches).

This differs from the old policy of adding the vanilla GNU/Linux version 
of a file to the trunk, which was primarily geared toward Debian users. 
This policy is outdated because there is now an OPENSSI-DEBIAN branch. 
Debian-specific files should only go there, and Red Hat-specific files 
should only go on the OPENSSI-RH branch. All distro-neutral files should 
be added to the trunk and aliased onto both branches.

Hopefully this clarifies things,

Brian

[CI] Away for two weeks

[Brian J. W. <Bri...@hp...>]

I'll be on vacation in Europe. I won't be responding to any e-mails 
until I return on Nov. 4.

Till then,

Brian

[CI] Re: OpenSSI running in RedHat 9.0 (With Further Information)

[Andres T. (CHASQUE) <an...@ch...>]

Brian:
=09In my company this is a test project, so we found a bounch of
computers that nobody was using. In one of them we've made a fresh
install of RH 9.0, then patched it with the latest packages avalible
from redhat and installed the SSI suite RPMs. While running install
we've found out that the the SysVinit complained, so we installed it
with --old-package. We also added libxml2. We also commented from
/etc/rc.sysinit the initialization of USB devices (redhat should make
this optional).

=09I've noticed that node #1 while booting complains as
follows: INIT:Error create initstate, failover compromised. Yesterday
we found a new computer for our experiment to act as node#2, we
configured it and made it boot from a floppy created in rom-o-matic.
This computer boots, and somewhere it stales at the beggining of the
init process. I hope to find some spare time to look after this in a
while.

=09Do yo have any idea of the next release date? I don't find my
self with the time and knowledge to fix my problems, but I can
betatest and return feedback.

=09=09=09Andr=E9s


 On Mon, 6 Oct 2003, Brian J. Watson wrote:
> Andres Tarallo (CHASQUE) wrote:
> > I'm running with success OpenSSI in a fresh redhat 9.0 instalation,
> > from the RPMs. Very little has to be done to achieve this goal.
> >=20
> > I'm clearing my notes (this i s atest instalation) and then I will
> > send a full report to let you know what differs from the standard RPM
> > instalation.
>=20
> I tried that myself with OpenSSI 0.9.9. One problem I ran into is that=20
> cluster_mkinitrd grabs the LTS version of libc and several other=20
> libraries, which are incompatible with the 2.4.18-18.7.x_ssi_3ssi=20
> kernel. It also made me nervous that I was significantly backrevving=20
> several packages, which broke dependencies of some non-SSI packages.
>=20
> The next release will be based on Red Hat 9 packages, so it will be=20
> inherently compatible with that distro.
>=20
> Brian

Re: [CI] OpenSSI running in RedHat 9.0

[Brian J. W. <Bri...@hp...>]

Andres Tarallo (CHASQUE) wrote:
> I'm running with success OpenSSI in a fresh redhat 9.0 instalation,
> from the RPMs. Very little has to be done to achieve this goal.
> 
> I'm clearing my notes (this i s atest instalation) and then I will
> send a full report to let you know what differs from the standard RPM
> instalation.

I tried that myself with OpenSSI 0.9.9. One problem I ran into is that 
cluster_mkinitrd grabs the LTS version of libc and several other 
libraries, which are incompatible with the 2.4.18-18.7.x_ssi_3ssi 
kernel. It also made me nervous that I was significantly backrevving 
several packages, which broke dependencies of some non-SSI packages.

The next release will be based on Red Hat 9 packages, so it will be 
inherently compatible with that distro.

Brian

[CI] OpenSSI running in RedHat 9.0

[Andres T. (CHASQUE) <an...@ch...>]

I'm running with success OpenSSI in a fresh redhat 9.0 instalation,
from the RPMs. Very little has to be done to achieve this goal.

I'm clearing my notes (this i s atest instalation) and then I will
send a full report to let you know what differs from the standard RPM
instalation.

Lease apologize my english,=20

=09=09=09Andr=E9s Tarallo

[CI] Re: [SSI-devel] Disabled CVS checkins

[Brian J. W. <Bri...@hp...>]

Brian J. Watson wrote:
> I've temporarily disabled checkins to the CI and OpenSSI repositories 
> until I finish merging the Red Hat branch up to the RH 2.4.20-20.9 
> kernel code base (the latest kernel for RH9).
> 
> I'll re-enable checkins and send out another note when I'm done.

I've checked into the OPENSSI-RH branch the new CI and OpenSSI kernel 
source based on Red Hat's 2.4.20-20.9 kernel. It currently builds and 
boots, but there may be a few bugs that still need to be tracked down. 
Note that you need to rebuild cluster-tools, since the system call 
numbers were changed.

There's a good chance I broke the kernel on the trunk, due to many of 
the CI/OpenSSI kernel files being COMMON. To help clean up this mess, I 
tagged both my CI and OpenSSI kernel sandboxes with PRE-RH9 before I did 
my checkins. We can branch any of these files if we need to.

An RH9 based release of OpenSSI will be available soon. It will support 
the GRUB bootloader rather than LILO. It will include an enhanced 
mkinitrd, instead of the old cluster_mkinitrd. Its support for NFS will 
be greatly enhanced by Dave. Various other enhancements and bugfixes 
will also be included. The release will also include some crude 
instructions for upgrading from the RH8-based 0.9.9 release to the new 
RH9-based release.

The repositories are once again open for business, and commit logging 
has been re-enabled.

Enjoy,

Brian

[CI] Re: [SSI-devel] Re:Disabled CVS checkins

[Aneesh K. K.V <ane...@di...>]

On Tue, 2003-09-30 at 11:57, Watson, Brian J. (HP) wrote:
> Aneesh Kumar K.V wrote:
> > Does that means the TRUNK  is going to 2.4.20  ?  
> 
> Only if someone else is willing to do that merge. I can offer some 
> instructions, but I don't have time to do much more than that.
> 


I will do that once you finish moving the redhat branch to 2.4.20-20.9.
I will send you the patch once i finish the merging . I don't want to do
the commit, because it is getting more and more complex with CVS. Is
that ok with you ?  


-aneesh 

Re: [CI] Re: Disabled CVS checkins

[Brian J. W. <Bri...@hp...>]

Aneesh Kumar K.V wrote:
> Does that means the TRUNK  is going to 2.4.20  ?  

Only if someone else is willing to do that merge. I can offer some 
instructions, but I don't have time to do much more than that.

Brian

[CI] Re: Disabled CVS checkins

[Aneesh K. K.V <ane...@di...>]

On Tue, 2003-09-30 at 02:28, Watson, Brian J. (HP) wrote:
> I've temporarily disabled checkins to the CI and OpenSSI repositories 
> until I finish merging the Red Hat branch up to the RH 2.4.20-20.9 
> kernel code base (the latest kernel for RH9).
> 

Does that means the TRUNK  is going to 2.4.20  ?  

In that case i may need to verify all the Alpha related changes.  

-aneesh 

[CI] Disabled CVS checkins

[Brian J. W. <Bri...@hp...>]

I've temporarily disabled checkins to the CI and OpenSSI repositories 
until I finish merging the Red Hat branch up to the RH 2.4.20-20.9 
kernel code base (the latest kernel for RH9).

I'll re-enable checkins and send out another note when I'm done.

Brian

Re: [CI] CI newbie question

[Brian J. W. <Bri...@hp...>]

Zou, Yixiong wrote:
> Hi 
> 
> I just downloaded CI recently and looked at the cluster API interface.
> I have couple questions that I wish you guys can answer for me. 
> 
> 1) Is there plans to port CI's kernel patch to the latest 2.6 kernel?

Sometime in the future it will ported to 2.6 along with the OpenSSI code 
that depends on it.

> 2) Is there plans to submit CI's kernel patch to the main-line kernel? 

It would be good if both CI and OpenSSI became part of the main-line 
kernel, but there are no specific plans right now. CI would be an easier 
sell, because it has far less hooks into other kernel code. 
Nevertheless, I think it needs to have a reasonably sized user community 
before Linus would accept it.

> 3) Is there any distro who currently ships CI? 

No.

> 4) Any production installations of CI you guys know of?

CI requires software that uses its features to truly be useful. By 
itself, it's just keeps track of cluster membership, albeit rather well. 
The only software I'm aware of that uses CI is OpenSSI.

> 5) Have you guys looked into the SA-Forum's AIS spec? Is there a
>    plan to be SA compliant? This question is a little odd, I sincerely
>    whish you guys don't get offended by this. I just want to know where
>    this project is going next. There are so many cluster solutions
>    out there and it will be very nice to have a common interface 
>    to users can easily integrate them together. 

I don't think any of us have looked at the AIS spec in any detail. It 
would be nice to have a common clustering API. The reason CI was 
spun-off from OpenSSI was to try to promote a common membership 
interface for all clustering software, but it's been difficult to 
attract other developers to build apps on it.

Hope this helps,

Brian

[CI] CI newbie question

[Zou, Y. <yix...@in...>]

Hi=20

I just downloaded CI recently and looked at the cluster API interface.
I have couple questions that I wish you guys can answer for me.=20

1) Is there plans to port CI's kernel patch to the latest 2.6 kernel?

2) Is there plans to submit CI's kernel patch to the main-line kernel?=20

3) Is there any distro who currently ships CI?=20

4) Any production installations of CI you guys know of?

5) Have you guys looked into the SA-Forum's AIS spec? Is there a
   plan to be SA compliant? This question is a little odd, I sincerely
   whish you guys don't get offended by this. I just want to know where
   this project is going next. There are so many cluster solutions
   out there and it will be very nice to have a common interface=20
   to users can easily integrate them together.=20

Thank you very much for your help. Best regards.=20

------------------------------------------------------------------------
Yixiong Zou (yix...@in...)
(503) 677-4988

All views expressed in this email are those of the individual sender. =20