mirrored plexes resync triggers

[KNR]

Hello,

I would like to understand what triggers mirrored volume plexes to resync (RDSRBACK) and how to skip this, incase if needed.  Particularly after good no.of days of update, if we reboot the box all the mirrored volumes goes to  SYNC state. I understand this is designed for good reasons. However, i would like to know how to skip it.

Thanks in advance.

Nagesh.

[Marianne]

This should not happen if a proper shutdown is done when all in-memory data is written to disc during a 'sync' operation.

Mirror resync is normally associated with system crash when writes to mirrored volumes were incomplete when the system went down.

Enable 'dirty region logging' to speed up resync of volumes.

If we know your SF version, we can point you to the relevant vxvm manual.
Else, go to https://sort.symantec.com/documents and select the SF version for your OS. 

[KNR]

Thanks for the advice on using DRL log. Most of the boxes I worked are running with solaris 10 + vxvm 5.x. I can say, it is *NOT* the case of improper shutdown which triggered resync. I patch 10-15 servers in a week and all the servers with 100+ days of uptime, this behaviour is observed. Curious if anyone has opened the vxrecover startup script code. Thanks in advance.

[Gaurav_S · Accepted solution]

Hi Nageswara,

Are you using Veritas Smartmove feature ? Can you paste the contents of  /etc/default/vxsf  ?

If a sync is getting triggered automatically by vxvm that would indicate for sure that something is got going for a clean shutdown of volumes ..

As you are doing number of activities per week .. . can you try following in one of patching

1. Stop volumes (vxvol -g <dg> stopall)

2. Deport the diskgroup (vxdg deport <dg>)

then go for rebooting your server, would be worth to know if still vxvm calls for a full sync. Be careful if you are using cluster on this server else cluster will anyways start all volumes configured in there.

when your system starts, it would be calling for vxvol startall & if in case the startall returns any error because of some volume or sub component is in inconsistent state, vxrecover will be called in which will trigger sync.

Also, it would be worth to know if any other customized scripts exist in SMF or rc directories which is caling for vxrecover ?  This can happen if snapshots are in use & someone intentionally calling sync to ensure snapshots are in sync ?

G