Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Sun, Jul 17, 2016 at 11:01:50PM +0200, Roland Kammerer wrote: > On Sat, Jul 16, 2016 at 11:34:41AM -0700, Jason Dillon wrote: > > > > FTR I’m using just the standard gcloud ubuntu 16.04 image: > > > > FTR I tried this a bunch of times and it worked as expected once or > > twice, then freaked out. :-( > > Hi Jason, > > so far I have not been able to reproduce it in my xenial VMs, but we > will investigate that ASAP. For now, thanks for reporting! Hi Jason, still no luck with reproducing (and from the trace it looks like the bug is below DRBD). Could you help us to reproduce it? Make sure that there are no drbdmanage resource and then start with a 200MB volume: "drbdmanage nv testres 200M -d2" Make sure there are no other resources, so the testres/volume should get /dev/drbd100. Then run that (sloppy) bash script[1], start with: ./test.sh 10 all You can trigger a mount/umount test only with: ./test.sh 10 notall I executed 50 test runs with "all" and 100 "mount/umount" without any issues. The only difference is see for now is that I execute it in my local xenial VMs and that my volume size might be much smaller. If that works, please remove the resource and create a new one with the size you started with originally (and adapt the "dd count"). Thanks, rck [1] ----------8<----------8<---------- #!/bin/bash for i in $(seq $1); do echo "Test: $i" if [ "$2" = "all" ]; then mkfs.ext4 -F -E discard /dev/drbd100 fi mount /dev/drbd100 /mnt if [ "$2" = "all" ]; then dd if=/dev/urandom of=/mnt/test.bin bs=1M count=150 fi umount /mnt dmesg | egrep BUG if [ "$?" = "0" ]; then echo "triggered bug" break fi done ----------8<----------8<----------