2.6.18-164.2.1.el5.028stab066.7 VMCORE [message #38218] |
Sun, 29 November 2009 17:27 |
Jean-Marc Pigeon
Messages: 27 Registered: October 2007
|
Junior Member |
|
|
Obviously there is a problem with 2.6.18-164.2.1.el5.028stab066.7, within 48 hours it crashed 3 times
(while 2.6.18-128.2.1.el5.028stab064.8 never give me even one trouble).
I activated KDUMP, such for the last crash generated a vmcore file.
Problem "crash" is not willing to give me any data, as there is no debugging information in the kernel...
I tried to recompile kernel with builddebug and buildkdump set to 1, but crash is still not happy...
Could somebody give me some hint how to extract meaningful data from vmcore and feed the list??
what is the best way to proceed?
Seems to me the two last crash occur at almost the same time
(may be a cron action trigerring fault?).
Many Thanks for your help.
|
|
|
|
|
|
|
|
|
|
|
|
Re: 2.6.18-164.2.1.el5.028stab066.7 VMCORE [message #38402 is a reply to message #38398] |
Tue, 15 December 2009 09:29 |
khorenko
Messages: 533 Registered: January 2006 Location: Moscow, Russia
|
Senior Member |
|
|
Hi,
Quote: | Is this patch the same as the bug fix for an NFS OOPS that was included in 028stab066.10 that was released a few days ago or different things?
|
No, the bug was found only yesterday and was not included into 66.10, which was released earlier.
Quote: | My main concern was the kernel provided without
debuginfo capabilities, such we (we as the users) were
not able to provide meaningful data (just guessing) to
fix/upgrade openVZ kernel.
|
Well, on one hand you are right, having a kernel built with debuginfo could help you provide us info about the problem.
But on the other hand:
1) most time memory dumps are redundant to findout a bug triggered an oops, complete oops messages are often enough.
2) At the same time configuring serial or network console is much simpler task than configuring crashdump, then use "crash" to provide necessary info - thus in most cases we ask for console logs.
3) Next: if a guy is advanced enough to configure crashdump and he is aware how to use "crash" utility, i believe it won't be so big problem for him to compile additionally a kernel with debuginfo enabled.
Jean-Marc, good luck with "vzgot"! Hope you'll share it with the community once it is ready!
--
Konstantin
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|