Bug #859: Ceph does not pass fsstress
osd spinning during fsstress
During some fsstress runs, I've managed to end up with an OSD spinning at 100% cpu, making no apparent progress, which the monitor thinks is down. This is running on kai with btrfs. The first time I thought it dismissed it as an unfortunate interaction between cfuse (running in my source dir) and the OSD (running in a different part of my source dir, but on the same btrfs partition).
Then I reproduced with the cfuse mount in a completely separate drive.
#1 Updated by Samuel Just almost 9 years ago
This seems to be caused by the following chain:
1) bufferlist iterator copy from a bufferlist containing one empty bufferpointer loops forever
2) filestore::read on an empty file seems to result in a bufferlist containing one empty bufferptr
TMAPUP op calls 1) with 2) since iterator to the beginning of 2) is not equal to end().
Will fix on monday.