Video Screencast Help
Symantec to Separate Into Two Focused, Industry-Leading Technology Companies. Learn more.

Flashbackup on RHEL fails with error 156

Created: 11 Oct 2012 • Updated: 31 Oct 2012 | 6 comments
This issue has been solved. See solution.

Hi,

VxSF 5.1 on RHEL5. Using vxfs snapshot method works perfect if only one raw device is in the directive list. the vxcache setting is configured in the snapshot options with a 100GB cache LUN, so I know space is not the issue. 2 seperate DG's, one vol one each, one FS on each. If I only enter one at a time in the backup selections, they both rip off there no problems. We want to have them both spin off serially, or possibly using multiple streams. I've tried multiple streams using the new stream directives, but as nothing works. I don't see this as a volume set limitation as they are totally independent. And, there is NO IO on the filessytem when trying to back them up, so the cache disk can't possibly be filling. I know we did this years ago in 4.x, but I can't seem to get this working.

I'll attach the bpfis later, but I don't see the clue in there, just that the snapshot could not occur.

thanks

Comments 6 CommentsJump to latest comment

Nicolai's picture

Have you tried separating the two rawdevice/cache devices by the NEW_STREAM directive

e.g.. (syntax below is not correct - it just intended for clarification)

/dev/dsk/vx/dgxx/lvolxxx

cache=cachevolxx

NEW_STREAM

/dev/dsk/vx/dgyy/lvolyy

cache=cachevolyy

Assumption is the mother of all mess ups.

If this post answered your'e qustion -  Please mark as a soloution.

schrammd's picture

Yes, every combination I can think of.

NEW_STREAM
/dev/vx/rdsk/davelrsdevappdg/lrs
cache=/dev/vx/rdsk/vx_cache/vx_cachevol
NEW_STREAM
/dev/vx/rdsk/davelrsdevdatdg/lrsdata
cache=/dev/vx/rdsk/vx_cache/vx_cachevol

This fails with "1: (69) invalid filelist specification"
If I put the cache setting above the rawdisk says the same thing.

I have 2 environments to test this on, one is still running 6.5.6, the other 7.5 and both fail the same way.

thanks

schrammd's picture

Is there some unwritten requirement that every raw device in the list must have a dedicated vxcache volume?

In the old days, at least with Solaris, one cache setting at the snapshot option level was enough to have 10 filesystems in the list seperated by nothing, and all 10 would sequentially fire off just fine. But, it seems the cool things ever done with Solaris were not ported to Linux.

thanks

schrammd's picture

That was indeed the missing piece. Apparently that was left out of the snapshot admin guide. One dedicated cache vol per FS in the list. Or so it appears. The other trick is to deselect the "perform snapshot backup" option. Ugh

SOLUTION
schrammd's picture

There is some algorithm which determines the size of the vxcache being used for the volume backing up. With very large volumes, the cache volume must be bigger as well although we can find no documen which says "for this size disk use this size cache". Oh well, it's working now.

thanks