dmesg error

[joe_91]

This is a 11.11 box and my SAP user gets segmentation error when he attempts to execute the sapdba tool. But the strange thing on the box is when i execute dmesg
(msgcnt 4061 vxfs: mesg 001: vx_nospace - /dev/vg04/SAP file system full (1 block extent)

but bdf is fine it shows that lvol is 60% used. so i assume it should(the segmentation error) should not 'coz of the dmesg error but something else. is that assumption correct? also on the syslog i see the same no space error 15 days back, but in dmesg it is one which conmes up last

Thanks

Joe

[Bill Hassell]

dmesg does not timestamp it's errors. If that is the last error in dmesg, it could have occurred 2 minutes ago or 20 hours ago. To timestamp dmesg items, use a cron job like this:

0,10,20,30,40,50 * * * * /usr/bin/dmesg - >> /var/adm/dmesg.log

Now some flavors of Unix use the name messages.log but it seems more appropriate to name the file similar to the process that creates it. Once this is running in cron, each new batch of entries will have a timestamp.

[A. Clay Stephenson]

I doubt that the segmentation (SIGSEGV?) violation has anything to do with a full filesystem although that should cause a more or less graceful program error. dmesg is simply a circular fixed-size buffer and its main value is that it is a place to write error messages when perhaps no other resource is available. Not all events which are logged in syslog are also logged in dmesg so the two are only distantly related. IN any event, this filesystem ran out of space at some point and the offending process probably died and the space was returned to the filesystem. You really need to use a debugger to examine the core file but almost certainly you need to contact SAP for help. It is possible that your error resulted from one of the kernel limits (maxdsiz, maxssiz) being hit but SAP support should be able to help there as well.