RPC Timeouts solution..

[brian_31]

Hi:
We are facing peiodic rpc timeouts and hence failovers(MCSG). There was a suggestion to shorten resolver timeouts by adding retrans and retries option as follows in resolv.conf..
retrans 1000
retry 2
what would be the side effects of these changes?? Pl. help

Thanks
Brian

[Sanjay_6]

Hi Brian,

How is rpc timeout affecting your SG cluster. you can increase the node timeout parameter under the cluster configuration. Do remeber if you change this parameter you'll have to reapply the cluster configuration.

Are you using the resolver to resolve the hostname and ip addess of the cluster nodes. If so you can use the file /etc/cmcluster/cmclnodelist to do so. You may have to create this file if it does not exist. Also add the SG node names and ip addess to the local host file.

Can you post the section of /var/adm/syslog/syslog.log file that has this timrout and sg failover information. You can do a cmscancl to gather the current SG cluster infomrmation.

Hope this helps.

Rggds

[Sanjay_6]

Hi Brian,

I'm sorry, but there was a mistake in my previous post. The file /etc/cmcluster/cmclnodelist is used as a substitute for .rhosts and not as a substitute for the resolver.

Sorry abt the wrong post.

Regards
Sanjay

[brian_31]

Hi Sanjay:

It is not an MCSG issue. These dns resolver timeout reduction was HP recommendation.

Thanks
Brian.

[Sanjay_6]

Hi Brian,

Maybe this link can help,

http://us-support3.external.hp.com/cki/bin/doc.pl/sid=a79eea281b2dd2b5ae/screen=ckiDisplayDocument?docId=200000059423940

Hope this helps.

Regds

[Sanjay_6]

Hi Brian,

Here is something more on REsolver timeout,

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B2355-90741/B2355-90741_top.html&con=/hpux/onlinedocs/B2355-90741/00/00/31-con.html&toc=/hpux/onlinedocs/B2355-90741/00/00/31-toc.html&searchterms=timeout%7cresolver&queryid=20020327-202525

Hope this helps.

Regds

[Giri Sekar. · Accepted solution]

Hi:

Looks like you have NFS problem. You can set the parameters RES_RETRY and RES_RETRANS to your nfs.mon script thereby avoiding altering the systemwide resolver timeouts. Also consider using files first(in case you have dns first)in your /etc/nsswitch.conf.Looks like your current setting is trying to reduce the timeout from 75 to around 10 to 12 seconds systemwide.

Thanks
Giri Sekar.

[brian_31]

Hi Giri:

We do have nfs.mon script and i did not know i can set these variables in nfs.mon. I will speak to the team and try to go for this as this will not affect the systemwide resolver timeouts.

Thanks
Brian.