3 Nodes Veritas Cluster Preventive Maintenance

[chadkazuki]

Hi,

Good day to you guys, may i have inquiry regarding Veritas Clustering. Sorry im novice for this clustering system. Here are the details,

I need to upgrade my powerpath version installed to my 3 HP-UX servers with Veritas Clustering configuration and running an oracle instance.

The installation will required of server rebooting, may i need to know the best practice how to reboot this servers? what services i need to stop

first prior to server rebooting?

Hoping for your quick response.

Thanks in advance,

Regards,

[mikebounds]

Depends on what you want to happen to the services running on the node you are rebooting:

1. If you want services to remain online, then run "hastop -local -evaucate" and this switches service groups to alternatate nodes (or you can switch service groups manually and then just run hastop -local)
2. If you want services to remain offline (particulary if you are rebooting all 3 servers at the same time), then run "hastop -local" and this offlines service groups and leaves them offline (or you can offline service groups manually and then just run hastop -local)

If you do a clean reboot (i.e like using "shutdown" command) then the O/S rc stop scripts normally do 1, but I always like to do it manually so I can confirm services online ok on other nodes.

Mike

[Gaurav_S]

I would like to suggest to check the new powerpath version you are trying to install is compatible with veritas volume manager or not ...

each of powerpath versions do have a compatibility with volume manager ... what is SF version you are using ? if you are 3.5 on HP 11.11 , do check compatibility....

Also, are you using CVM/CFS ? is it a parallel cluster ?

also as Mike said .... first question is, what is your plan of upgrading powerpath ? one node at a time or u have outage to do all the 3 nodes at the same time ??

if former is the answer, switch over the service groups to other nodes (hagrp -switch), stop the cluster (hastop -local) or use the stop-evaculate option as suggested by Mike.... this will leave cluster runing with 2 nodes & you can carry on with upgrade on 3rd node .. once upgrade done & rebooted the server, server will automatically join the cluster ... you can then repeat the procedure for other servers ..

Gaurav

[chadkazuki]

Hi guys thanks for the inputs.

I think i'm going to install powerpath one at a time for the three servers. Is this okay if I'll stop or execute the "hastop -local" cmd for the 1st and 2nd server? then I will perform the maintenance such as installing and server rebooting. Then after i finished for the 1st & 2nd server i will start both service ("hastart") and stop the 3rd server "hastop -local" then continue for the maintenance until I finished.

We're planning the upgrade because of our vendor remediation, and related to capacity upgrade on our machine. They told us that our current version installed to our machine is not supported anymore, that's why
we're upgrading the powerpath.

@Gaurav
with regards to compatibility of the powerpath version, our vendor gave us the remediation for the possible software/firmware to be installed to our machine,i guess they knew the compatibility issue.

For the SF version sorry i'm new for this, how i check this? is this the same of uname -a? is that so here's the output "HP-UX PRNDPD01 B.11.31 U ia64 2642932093 unlimited-user license"

For the CVM/CFS, please help me to check what i'm using.
For the parallel cluster, i think yes. because all 3 servers is ONLINE.

Regards,

Chad

[Gaurav_S]

So for the steps you are using, approach is fine .. no issues if you take down 2 servers at the same time (I would have preferred 1 server first ... if successful, go ahead with other two) ..

Please note, "hastop -local" should be executed from the node where you want to stop the cluster ....

When you say your vendor has suggested to upgrade, who is the vendor ? is that EMC ?  If yes, then I assume EMC would have taken care of compatibilty with vxvm version .... if its some service provider, then still double check from release notes ..

your HP version is 11.31 so I assume that vxvm will be updated one (5.x)

check "   swlist | egrep 'vxvm|vcs|gab|vxfen'    "

for CVM, paste this output

# hastatus -sum

# gabconfig -a

If you seem to have a "CVM" service group in above output which is online on all the nodes at the same time, then lyou are using cvm ....

Gaurav

Gaurav

[mikebounds]

I agree with Gaurav, I would upgrade one server first, especially if node 3 is normally an inactive node (doesn't run service groups).

CVM/CFS is a parallel cluster when you concurrently mount filesystems on servers, so the same service group containing CFSMounts is online on more than one system at the same time and you would also have a service group called cvm.

Output of "hastatus -sum" would help.

Mike

[chadkazuki]

Hi Gaurav,

Thanks for your advise, I should go to node1 first. then if successful, go to remaining nodes.
Your right also our vendor is EMC, they are the one who gave us the remediation.

btw, here's the output
# swlist | egrep 'vxvm|vcs|gab|vxfen'
  VRTSgab                               5.0.31.0       Veritas Group Membership and Atomic Broadcast by Symantec
  VRTSvcs                               5.0.31.0       Veritas Cluster Server by Symantec
  VRTSvcsag                             5.0.31.0       Veritas Cluster Server Bundled Agents by Symantec
  VRTSvcsdc                             5.0.31.0       User Documentation for Veritas Cluster Server by Symantec
  VRTSvcsmg                             5.0.31.0       Veritas Cluster Server Message Catalogs by Symantec
  VRTSvcsmn                             5.0.31.0       Manual Pages for Veritas Cluster Server by Symantec
  VRTSvcsor                             5.0.31.0       Veritas High Availability Agent for Oracle by Symantec.
  VRTSvcsvr                             5.0.31.0       VERITAS Volume Replicator Agent for VCS
  VRTSvxfen                             5.0.31.0       Veritas Fencing by Symantec
  VRTSvxvm                              5.0.31.1       Veritas Volume Manager by Symantec

Thanks also with you mike, i appreciated all inputs.
 

Regards,

Chad

[mikebounds]

I believe this is version 5.0MP3 RP1.  Can't tell if you have CVM from this, but output of "hastatus -sum" would confirm that you don't have CVM.

Mike