10:36 p.m.
Dear All,
I've seen several cases of crashed client machines lately. There's a
high load on the lustre servers and the clients seem to run into
time-outs and get evicted. My clients run Lustre 1.8.9 and the servers
have 2.1.3, 1.8.7, 1.8.8.
Is there a known bug in the 1.8.9 client that could cause this?
Regards, Götz Waschk
12:48 a.m.
On 10/05/13 13:36, Götz Waschk wrote:
Dear All,
I've seen several cases of crashed client machines lately. There's a
high load on the lustre servers and the clients seem to run into
time-outs and get evicted. My clients run Lustre 1.8.9 and the servers
have 2.1.3, 1.8.7, 1.8.8.
Is there a known bug in the 1.8.9 client that could cause this?
I ran into a problem with 1.8.9 patchless clients where once the client
lost communication with the OSS, it didn't recover.
https://jira.hpdd.intel.com/browse/LU-3067
That shows up as an LBUG on the client (you've only posted the server
log). I've rolled back to 1.8.8 clients.
Chris
1:33 a.m.
On Fri, May 10, 2013 at 4:48 PM, Christopher J. Walker
<C.J.Walker(a)qmul.ac.uk> wrote:
I ran into a problem with 1.8.9 patchless clients where once the
client
lost communication with the OSS, it didn't recover.
https://jira.hpdd.intel.com/browse/LU-3067
That shows up as an LBUG on the client (you've only posted the server
log). I've rolled back to 1.8.8 clients.
Dear Chris,
I guess my problem is different as there wasn't this LBUG on my
client. My log sample contained the server messages and the ones from
the client (bladeed) that have reached the log server.
Regards, Götz
6:14 a.m.
I guess my problem is different as there wasn't this LBUG on my
client.
Are you sure? The LBUG message probably just didn't make it to the syslog server.
Do you have a serial or fake-vga console? -> any messages there?
How are you logging to the syslog server?
Kernel panics won't make it to the logserver with an userspace syslog daemon.
You'll have *much* better chances capturing a kernel panic by using netconsole.ko
Regards,
Adrian
--
RFC 1925:
(11) Every old idea will be proposed again with a different name and
a different presentation, regardless of whether it works.
11:02 p.m.
Hi,
On Fri, May 10, 2013 at 10:14 PM, Adrian Ulrich <adrian(a)blinkenlights.ch> wrote:
Are you sure? The LBUG message probably just didn't make it to
the syslog server.
you are right, the syslog daemon won't have submitted these
messages.
I'll attach the messages from the virtual consoles of several crashed
clients.
Looks like there might have been an LBUG after all.
Regards, Götz
________________________________
3416
days inactive
3418
days old
4 comments
3 participants
participants (3)
-
Adrian Ulrich -
Christopher J. Walker -
Götz Waschk