CentOS 5.6, mail server worked perfectly for months but ...
I'm using centos since CentOS 5.2, always upgraded it with yum update without any problem but leaving the same 2.6.18 kernel from linode.
Yesterday I decided to switch to the new 2.6.38 paravirt and rebooted my server.
It seems that all worked out fine until night when my server started loosing emails, I lost all the email that they sent me in the night.
What is the possible reason of this problem?
Thanks.
7 Replies
The cron daily performs very few action like a dump of the database, sent the admin an email saying him the size of a file (important for us), and sync the time with ntp.
What do you think that have created this problem on the latest paravirt kernel?
As I saied with the latest legacy 2.6.18 we haven't got any problem for months, with the latest paravirt this problem happened as soon as cron.daily has been executed.
Is there anything interesting in the logs?
What have you tried to do to fix it?
@hoopycat:
Which mail server software are you using?
Is there anything interesting in the logs?
What have you tried to do to fix it?
I have re-executed the cron.daily to see if this was the cause since the problem started at the same hour that I execute the daily cron, now it seems to work.
From my dashboards I noticed high CPU usage from the 4AM to 6AM, I stopped receiving emails at 4AM.
In the logs I cannot see any interesting and I don't know how to solve it.
What do you think?
–------------------- Kernel Begin ------------------------
WARNING: Kernel Errors Present
rtccmos: probe of rtccmos failed with error -38 …: 2 Time(s)
May 3 11:55:16 li62-51 postfix/qmgr[3686]: 98220104BE: to=<
tion refused)
netstat -ntlp | grep 25
And:
dig address.com mx
@hoopycat:
Hmm, what do you get for:
netstat -ntlp | grep 25
And:
dig address.com mx
it seems that 2.6.38 paravirt causes kernel panick on CentOS,
it is a known problem from the linode's team so I switched on the 2.6.35 that has other known problems like this:
but at least it should not panick.
I hope that the known problems will be solved soon, I will update thread from time to time to see if something is moved.