> Today we have experienced again the same problem but in this case > the computer which was under heavy load and locked was the > gluster client (not the server). Unfortunatelly there was > absolutely nothing in the kernel log, which makes me thing that > this is produced by > locks on some part of the disk system that freezes any > application trying to read or write to the disks. Just a small note: that's the kind of lock-ups I was seeing too, client induced, disks completely inaccessible, no error/dmesg log. I have now updated to RHEL 5.4 and I am lock free so far after heavy testing and lots of software building that was previously hanging my systems. I have also enabled magic-sysrq so I can get at least some info should the hangs re-occur. I would recommend everyone else with logless hangs to do the same if possible. Kudos to all those involved in this bug-hunt fest, it can only serve to make glusterfs stronger! Jeff.
