vx_multi_bufinval3

[Tom Danzig]

Recieved the following in the syslog this AM:

vx_multi_bufinval3: bcount 0x800 blkno 0x8455a flags 0x20808a nblks 0x1 ibsize 0x400

Server is a N4000 running 11.00 with OnineJFS. I searched on the forums and HP tech knowledge base and found nothing pertinent. Also could not find this error listed in the "HP JFS 3.3 and HP OnLineJFS 3.3
VERITAS File System 3.3 System Administrator’s Guide". Anyone have any idea what this message means and how serious it may be?

Thanks for any insight,
T

[Hemanth Gurunath Basrur]

Hi Tom,

I am not sure what the message means and how serious it may be.

In the foll. link,
http://archives.neohapsis.com/archives/hp/2003-q2/0032.html

I found that there is a patch.

PHKL_28966 s700_800 11.00 VxFS3.3 Data Page Fault in vx_multi_bufinval

I am not sure if this patch fixes your problem.

Please check.

HTH.

Regards,
Hemanth

[Cheryl Griffin]

The vx_multi_bufinval3 message indicates a buffer is about to be released. It is informational only, although the information is useless as it appears encrypted.

This message existed only at JFS 3.1 on 11.00. The same message does not exist in JFS 3.3.

In some cases where the message was reported, umount and fscking the filesystem, caused the messages to cease.

[Tom Danzig]

Hemanth,

I saw that patch (the only thing that came up at all in my search) and it fixes a data page fault for a very similar (but not identical) errror message. No panic occured here.


Cheryl,

Thanks for the info. I am on JFS 3.3; not 3.1 here (as far as I can tell anyway). Unfortunately, I cannot determine which file system the error message refers to so I cannot unmount and fsck. That will have to happen for all file systems at next reboot.

Hopefully this is an informational message only.

[Dave Olker]

Hi Tom,

I just checked the VxFS source code and I believe Cheryl is correct about this being a VxFS 3.1 message only. While VxFS 3.1 and 3.3 both contain the vx_multi_bufinval() routine, only the 3.1 version has the condition check that results in printing the message you're seeing.

In fact, the entire check is removed from the 3.3 version of vx_multi_bufinval, although that may simply be a result of redesigning this routine and the functions called by this routine, as opposed to Veritas deciding the check is bogus.

If you've installed VxFS 3.3 on this system, which would have been done separately since 11.0 didn't ship with 3.3, I'd be concerned about the fact that a VxFS 3.1 message is still being displayed. You may want to verify which VxFS version you're running, and if you believe 3.3 is supposed to be installed on this system then verify whether the installation completed thoroughly and successfully.

Regards,

Dave

[Tom Danzig]

Thanks Dave. I cannot seem to determine exactly what version of JFS I am using. swlist -v JournalFS shows:

data_model_revision 2.40

It does not seem to indicate a version. How do I get the JFS version from swlist?

So I can safely ignore the vx_multi_bufinval3 error message?

Thanks,
Tom

[Cheryl Griffin]

Run this command:
# /sbin/fs/vxfs/subtype
vxfs3.3
--> if 3.3 is installed.

The command will error or be blank, I believe, if you have 3.1.

[Tom Danzig]

Thanks Cheryl. This command is not on the server. Does this mean I'm at 3.1?

I have:
/sbin/fs/vxfs/bcheckrc
/sbin/fs/vxfs/extendfs
/sbin/fs/vxfs/fsck
/sbin/fs/vxfs/fsdb
/sbin/fs/vxfs/mkfs
/sbin/fs/vxfs/mount
/sbin/fs/vxfs/newfs
/sbin/fs/vxfs/umount

On another server, it is there and reporting vxfs3.3.

[Cheryl Griffin · Accepted solution]

yes - so 1 server has 3.1 and the others have 3.3.

Can you upgrade this one server to 3.3?

[Tom Danzig]

Thanks for the verification of 3.1. Upgrade at this time doubtful.

[Dave Olker]

Hi Tom,

Since upgrade can't be done, the next best thing would be to schedule a full fsck().

Every reported instance of this message has been met with the same recommendation from HP: it's probably safe to ignore the message, as it is likely just a debug statement that got left in the code by mistake, but just in case - schedule a full fsck as soon as time permits.

Regards,

Dave

[Tom Danzig]

Thanks Dave. The only issue is ... which file system is it? It's not identified in the error message. I'll wait for the next reboot.