scjody@clusterfs.com
2007-Jun-18 16:36 UTC
[Lustre-devel] [Bug 12762] New: Extreme slowness during sanity.sh
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=12762 I''m seeing extreme slowness during parts of sanity.sh on b1_6, approximately once every 5 runs. The first test where this is really apparent is 27m, specifically during the "lfs setstripe" operations. I have only been able to reproduce it during full acceptance-small runs (runtests then sanity) but that may be due to its intermittent nature. At Alex''s request, I reproduced the problem with mballoc off. I will attach "lctl dk" logs produced by a slight change to sanity.sh (I dump the log immediately after each "lfs setstripe" in test 27m. The only other thing I''ve noticed in dmesg is many instances of: LustreError: 1419:0:(lov_request.c:677:lov_update_create_set()) error creating fid 0x156 sub-object on OST idx 0/1: rc = -110 Sample slow run: == test 27m: create file while OST0 was full =================== 17:49:52 (11822 03392) dd: writing `/mnt/lustre/d27/f27m_1'': No space left on device 84242+0 records in 84241+0 records out PASS (1306s) Sample good run: == test 27m: create file while OST0 was full =================== 17:37:32 (11822 02652) dd: writing `/mnt/lustre/d27/f27m_1'': No space left on device 84246+0 records in 84245+0 records out PASS (22s)