regarding Service Guard patch PHSS_36898

[kumardev]

Hi,
I am abt to install the patch PHSS_36898, but in the instructions of the patch its recomended to halt the Service guard, but as my system is a production system so I wont be able to halt the service guard as my other packages are also running in the same cluster, so I was thinking of removing the node from the cluster, applying the patch on the node and then again re-joining the node to the cluster. Any suggestions, also in the instruction it says that I have to " recreate the package control script by using cmmakepkg and migrate all user customization from old control script to newly created" has anyone got any idea as to how to migrate the user custmization from the old pkg control file to the newly created file flawlessly??

What is the best approach for it?

Regards
kumar

[Duncan Edmonstone · Accepted solution]

kumar,

i) any reason why you aren't installing the more recent recommended patch PHSS_37608?

ii) The patch notes actually say "Halt Serviceguard on the node the patch is to be installed on." So there's no need to halt the whole cluster - just fail over all packges to other node(s) and then cmhaltnode the node you are going to patch. Remember that once you have applied a patch to one node in a cluster you should quickly arrange to get it applied to all others and until you have done this you should issue *all* serviceguard commands from the patched node.

iii) reading the notes on having to re-create the package control scripts, it seems to indicate this is only required for certain circumstances - have you read through all the notes carefully and confirmed you really need to do this?

HTH

Duncan

[Emil Velez]

Would doubt that you have to reapply the cluster after installing the patch. Serviceguard after being patched should be able to read and process the binary file.


move packages to other nodes
cmhaltnode -f (or this will do it)
install patch
cmrunnode

One of the basic tenants of serviceguard is rolling upgrades.

There are many patches for LVMv2, lvmv2 as a lock disk, Patches for toolkits.

lots of different patches.

[Johnson Punniyalingam]

Hi Kumar,

Its looks straight forward , Patch does'nt require reboot , You Just failover pkg and haltnode and swinstall the Patch, Can you also Install the recommended patch.


What is the best approach for it?


1) Halt Serviceguard on the node the patch is to be installed on.

You can just failover the pkg to another node

# cmhaltnode

2) Install this patch on that node.

# swinstall -s
3) Restart Serviceguard on that node.

# cmrunnode

failover the pkg again.

4) Patch needs to be installed on all nodes in the cluster.

You perform the same setps for other nodes in the Cluster.

Goodluck !!

Thanks,
Johnson

[kumardev]

Thanks to all for ur kind help.

As this patch is recently recomended by HP so I am installing this patch inspite of the more recent patch. One more thing, as my cluster holds several other pkgs, so after the installation of this patch I want to keep this node under monitoring for a few days,so during this monitoring period i wont be able to patch the other nodes in the cluster, do you think keeping other servers un-patched during the time of monitoring will affect the cluster functionality?? or I have to patch all the nodes of the cluster one after the other for all the cluster nodes at once as all the pkgs that are running are critical and that will be of a greater risk as well, any comments??

Regards
Kumar

[Johnson Punniyalingam]

Hi Kumar,

>>do you think keeping other servers un-patched during the time of monitoring will affect the cluster functionality?? <<<

Technically shouldn't "affect" but partically "touch and go"...

that's why "Devlopment " Server comes in handy" while these kind of Question raise..But what to do some Organisation "don't provide those comfort" no choice... :)

or I have to patch all the nodes of the cluster one after the other for all the cluster nodes at once as all the pkgs that are running are critical and that will be of a greater risk as well, any comments??

well , I d'not see any greater risk ... But you should make sure after monitoring, you need patch the other nodes.

if "down time " has been approved, I would take the downtime,

stop all the pkg runnning in node
halt all the node in the Cluster
patch for all the nodes in the Cluster,
And bring the Cluster.
(easy , But you need to afford downtime)

Thanks,
Johnson

[kumardev]

Many thanks for the replies....
Just one more thing,
1)Do i have to restart the cluster after applying patches to all the nodes of the cluster.
2)How the user customization that I have done on the pkg control files will be reflected in the new pkg cntl file? do I have to manually edit the file? what is the best approach for it??

Regards
kumar

[Johnson Punniyalingam]

Just one more thing,
1)Do i have to restart the cluster after applying patches to all the nodes of the cluster.

No .. Need

>>>>2)How the user customization that I have done on the pkg control files will be reflected in the new pkg cntl file? do I have to manually edit the file? what is the best approach for it??<<<<<<<,

cluster configuration file must be re-applied.Use the following steps to complete the changeover

Verify the configuration files and update the cluster binary file.
Example:

# cd /etc/cmcluster
# cmcheckconf -C cluster.ascii -P pkg1/pkg1.conf -P pkg2/pkg2.conf ...
# cmapplyconf -C cluster.ascii -P pkg1/pkg1.conf -P pkg2/pkg2.conf ...

man cmcheckconf & cmapply.conf for more information.

Thanks,
JOhnson