shared QFS, atime update changes mtime too

Greetings,

We have a small shared SAM-QFS setup here: Four 280R's, nine shared

filesystems and a few multi-reader/single-writer filesystems. We've

recently noticed that one of the client 280R's, when it reads a file on one

of the shared FS's, will update the mtime on the file it reads (so

mtime == atime == ctime).

None of the other client 280R's cause this to happen, nor does it happen

on the metadata server. The "problem" client does not cause this to

happen on any of its local-disk filesystems, nor on any of the other

shared FS's. There don't appear to be any other problems with this

filesystem -- the contents of the files seem OK, as do the directories,

permissions, etc.

All of the four servers appear to be running the same SAM-QFS 4.0 patch

revs (113172-08), and all seem to have consistent mcf and vfstab entries,

etc.

I've not yet found mention of mtime problems with my Googling,

SunSolve searching, etc., and am hoping someone in this forum

can point me in the direction of a solution, or has seen this behavior

themselves.

Thanks and regards,

Marion

[1220 byte] By [hakanson] at [2007-11-25 23:03:16]
# 1
Another data point: The problem went away after unmounting and remounting the problem filesystem on the problem client. No other changes were made. I doubt this is a "fix", though, because the problem has recurred at some point after a reboot, so it appears likely to return.
hakanson at 2007-7-5 17:55:10 > top of Java-index,Storage Forums,Storage General Discussion...
# 2

Further info: I found in the README for patch 113172-10 mention of what could be a similar problem. In the description for version -05 of the patch:

4838041 Shared QFS: modification dates keep getting changed on the client

and later, in version -07 of the patch:

4965143 Remove mod time fix

As I mentioned, we're running version -08 of this patch, and there does not appear to be (yet?) a fix mentioned for this issue.

Hmm, I should probably stop talking to myself....

Marion

hakanson at 2007-7-5 17:55:10 > top of Java-index,Storage Forums,Storage General Discussion...
# 3
I am follow the tread but have not had a chance to look at our QFS system. Has been crazy today.Are you verifiy this with the "sls -D" command or and sfind? We also are running -08, but we only have a single writer/reader and 3 readers.
error at 2007-7-5 17:55:10 > top of Java-index,Storage Forums,Storage General Discussion...
# 4

<table border="0" align="center" width="90%" cellpadding="3" cellspacing="1"><tr><td class="SmallText"><b>error wrote on Tue, 28 February 2006 14:32</b></td></tr><tr><td class="quote">

I am follow the tread but have not had a chance to look at our QFS system. Has been crazy today.

Are you verifiy this with the "sls -D" command or and sfind? We also are running -08, but we only have a single writer/reader and 3 readers.

</td></tr></table>

Yes, "sls -lD" shows access, changed, and modification times all the same (as last access); The creation, attributes, and residence times remain unchanged. As I said, unmounting and remounting the shared client that's having the problem causes the problem to go away (for how long, we don't yet know).

hakanson at 2007-7-5 17:55:10 > top of Java-index,Storage Forums,Storage General Discussion...