Jim,
there was a bug fix for the proto mismatch crash that came after
pvfs-2.8.1. Would you be willing to patch you 2.8.1?
https://trac.mcs.anl.gov/projects/pvfs/ticket/82
http://www.pvfs.org/fisheye/browse/PVFS/src/server/proto-error.sm?r1=1.10&r2=1.11
kevin
On Jul 22, 2009, at 3:45 PM, Jim Kusznir wrote:
Hello:
I am working on upgrading my pvfs installation from 2.7.1 to 2.8.1
today, and ran into a problem. When I started the 2.8.1 server for
the first time, it started working away on a conversion. It
eventually finished the conversion, but then crashed when I tried to
use it. Here's the output from the first of 3 pvfs servers (all my
servers are dedicated metadata and i/o servers. The url from all my
clients points to this server):
[D 07/22 08:43] PVFS2 Server version 2.8.1 starting.
[E 07/22 08:52] Trove Migration Started: Ver=0.1.3
[E 07/22 09:14] Trove Migration Complete: Ver=0.1.3
[E 07/22 09:14] Trove Version Set: 0.1.4
[E 07/22 09:35] Error: poorly formatted protocol message received.
[E 07/22 09:35] Protocol version mismatch: received major version 5
when expecting 6.
[E 07/22 09:35] Please verify your PVFS2 installation
[E 07/22 09:35] and make sure that the version is consistent.
[E 07/22 09:35] PVFS2 server: signal 11, faulty address is 0x10,
from 0x43f114
[E 07/22 09:35] [bt] /usr/sbin/pvfs2-server [0x43f114]
[E 07/22 09:35] [bt] /usr/sbin/pvfs2-server [0x43f114]
[E 07/22 09:35] [bt]
/usr/sbin/pvfs2-server(PINT_state_machine_invoke+0xcf) [0x44f61f]
[E 07/22 09:35] [bt] /usr/sbin/pvfs2-server [0x440608]
[E 07/22 09:35] [bt]
/usr/sbin/pvfs2-server(PINT_state_machine_invoke+0xcf) [0x44f61f]
[E 07/22 09:35] [bt]
/usr/sbin/pvfs2-server(PINT_state_machine_next+0xbc) [0x44f92c]
[E 07/22 09:35] [bt]
/usr/sbin/pvfs2-server(PINT_state_machine_continue+0x1e) [0x44f4ae]
[E 07/22 09:35] [bt] /usr/sbin/pvfs2-server(main+0xa7e) [0x413b5e]
[E 07/22 09:35] [bt] /lib64/libc.so.6(__libc_start_main+0xf4)
[0x33d181d8a4]
[E 07/22 09:35] [bt] /usr/sbin/pvfs2-server [0x410b49]
[D 07/22 13:37] PVFS2 Server version 2.8.1 starting.
At the last line, I restarted my server having found it broke, and my
system load went up, with pvfs2 using slightly over 100% of a core of
CPU time (like it did in the conversion),and its not responding to any
requests.
All my clients have been upgraded to the same version (2.8.1), so I'm
not sure where the old version request came from.
--Jim
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
