Re: migrate user home from Solaris to RHEL 3AS
- From: Georg Klein <gk@xxxxxxxxxx>
- Date: Mon, 25 Sep 2006 07:01:20 +0000 (UTC)
grodenhiATgmailDOTcom <grodenhi@xxxxxxxxx> wrote:
We have a RHEL 3AS that is mysteriously crashing with no decernable
pattern. It was withstood VERY high memory/CPU load with no issue, but
then has crashed with no users logged in (hence very low load). The
machine will stay up for about 3-4 days then lock completely (nothing
in logs, can't even netdump), however it will respond to pings. The
only way to get it back to to pull power and restore it. We have moved
the drives to new hardware (essentaily ruling out bad proc/memory). I
have two possible theories:
What version of rhel3 you are running? We did experience some problems
when we upgraded to UP7 (especially to early installations of rhel3,
where we did not disable the startup of the audit-daemon). Even
unconfigured (ie. running with the defaults) audit-daemon produced
such a large amount of data, that the filesystem threshold was reached
and the audit-daemon suspended execution: with the result, that no login
was possible (looked like a complete lock of the machine in
combination to high load). However, the machine responded to pings. As
you have to, we had to hard reset the machine to bring it up again.
You may want to search for messages in the log files regarding to
audit-daemon: we found the above mentioned message 'auditd suspended
execution'. After disabling auditd startup, the problem were gone. You
may also want to have a look at the /var/log/audit.d/ directory, to
see if there are a lot of save.x files.
1.) After this machine was build we wholesale migrated about 100 user
home direcrtories from Solaris 8 to this machine. All of these users
home directories had dot-files for system and app setting (.gnome,
etc...). In theory anything that RHEL uses would over write the old
solaris dot files, is this a correct assumption? If not, would this
cause instability enough to crash the entire system?
Don't think that. We did migrate a lot of Solaris boxes to Linux
without any problems, even if there were a lot of users.
2.) The parent directory for all these home directories is a symbolic
link to another filesystem. So essentially anyone logging in is being
put into a symbolically linked home directory. Could this cause these
random crashes?
Again: I do not think that this causes the crashes.
This system was up for well over a month when it was only being used by
10 or so users, now it is being logged into and out of by 100+
different users off and on all day (perhaps about 30-40 users total on
the system at the same time). Each of these users is loading an
individual Gnome session (this server is running SunRay software
driving 40+ dumb Gnome terminals). Thanks for any ideas/suggestions in
advance!!!
Don't know whether that is the reason for your crashes, but a lot of
more users produce a lot of more audit entries - so this could be your
problem as it has been ours.
hth
Georg
.
- References:
- migrate user home from Solaris to RHEL 3AS
- From: grodenhiATgmailDOTcom
- migrate user home from Solaris to RHEL 3AS
- Prev by Date: Re: Samba on Redhat 9.0
- Next by Date: Re: dhcp ?
- Previous by thread: migrate user home from Solaris to RHEL 3AS
- Next by thread: migrate user home from Solaris to RHEL 3AS
- Index(es):
Relevant Pages
|