Description of problem: revolver has having fun picking off nodes and then bringing them back up once they had been recovered and on the 9th iteration it was trying to bring morph-05 back up and it tripped this assert when attempting to mount one of the filesystems: GFS: Assertion failed on line 359 of file /usr/src/cluster/gfs-kernel/src/gfs/recovery.c GFS: assertion: "!error" GFS: time = 1095875443 GFS: fsid=morph-cluster:corey0.4 Kernel panic: GFS: Record message above and reboot. How reproducible: Didn't try
Reproduced this one while running the same revolver senario. Again this assert/panic happened while attempting to mount the first filesystem on the recovered node.
Reassign
can you describe the setup? (nodes, fses, all that stuff.)
nodes: morph-01 - morph-05 fses: 3 - 5 mounted on all nodes I/O being run: genesis genesis -n 500 -d 50 -p 3 accordion accordion -p 3 accrdfile1 accrdfile2 accrdfile3 accrdfile4 accrdfile5 growfiles growfiles -i 0 -N 500 -n 3 -b iogen iogen -f buffered -m sequential -s read,write,readv,writev -t 1b -T 100000b 100000b:rwbuflarge | doio -avk iogen iogen -f sync -m sequential -s read,write,readv,writev -t 1b -T 100000b 100000b:rwsynclarge | doio -avk Nodes run I/O, a subset gets shot, they get brought back up, repeat :)
gulm or dlm?
dlm
I keep hitting these other three bugs trying to reproduce this one.
do you still see this bug?
Corey, have you seen this one? If not, can we close it for now?
Have not seen this bug in almost a year, will reopen if seen again.