[OpenAFS] Problems in the last 2 days
Klaas Hagemann
kerberos@northsailor.de
Wed, 29 Jan 2003 09:35:45 +0100
Hi,
i had lots of problems with my file-servers in the past few days and
posted a lot of, lets say not very "usefull" messages or #bullshit#.
Sorry for that and thanks to everyone who tried to help me.
No i thing i can discribe the error a bit better:
Bevor thursday i had 2 Management-Server and 1 Fileserver. The
Management Server only hostet read-only replicas for the subddirectorys.
All the Volumes for the home-directories (20000) were at the one
fileserver. The system was running very stable with this configuration.
Thursday evening i moved ca.500 Volumes with User Home Directories to
another, new Fileserver using the vos move command.
After that i got lots of problems with the processes on my fileserver.
First i used the pthread fileserver. There the fileserver-processes
simply stopped working from time to time, so that the volumes were not
reachable any more.
The volserver kept working, so that "vos examine >volume<" gives a
successfull return. The System could not be shutdown any more and the
processes had to be killed by hand.
Than i switched over to the PWD fileserver. There the fileserver process
itself works fine, but after 5-6 hours the kernel was not able to
allocate any more memory. So the whole system crashed and had to be
rebooted over the "reset-button".
Another error was that the volserver stopped working but the fileserver
were still running. So "vos examine >volume<" delievered a failure but
the volume still was reachable.
Then i moved the whole volumes back to the first file server and now
everything works stable again. I am very sure to be able to reproduce
this error, because nothing else happend on the network.
What may cause this problem?
I thought of problems in the synchronization of the 2 database
management servers, are there any known problems when moving lots of
volumes?
Please let me know if you need any further information.
I use openafs-1.2.7 on suse linux 7.3, /vicepa is on ext3 with lvm.
Thanks
Klaas