Red Hat Bugzilla – Bug 800779
clear-locks should not trigger self-heal of files
Last modified: 2015-11-03 18:04:14 EST
Created attachment 568195 [details]
clear-locks log file
Description of problem:
when clear-locks on a file is executed it performs lookup on the file specified. if lookup detects any pending operations, it triggers self-heal along with clearing locks if any locks were head on that file.
Ideally clear-locks should unlock any (posix/inode/entry) locks held on a file and should not trigger self-heal.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.create a replicate volume. start the volume
2.create 2 gluster mounts from client1
3.execute 'dd if=/dev/zero of=gfsf1 bs=1M count=10240' from mount1
4.execute 'ping_pong -rw gfsf1 2' from mount2
5.execute 'gluster volume clear-locks datastore /gfsf2 kind blocked posix' from one of the server in storage pool
[2012-03-07 18:31:15.509693] D [afr-common.c:1137:afr_lookup_set_self_heal_params_by_xattr] 0-datastore-replicate-1: data self-heal is pending for /gfsf2.
[2012-03-07 18:31:15.509730] D [afr-common.c:1137:afr_lookup_set_self_heal_params_by_xattr] 0-datastore-replicate-1: data self-heal is pending for /gfsf2.
[2012-03-07 18:31:15.509814] I [afr-common.c:1313:afr_launch_self_heal] 0-datastore-replicate-1: background data self-heal triggered. path: /gfsf2, reason: lookup detected pending operations
[2012-03-07 18:31:15.509889] D [afr-self-heal-metadata.c:78:afr_sh_metadata_done] 0-datastore-replicate-1: proceeding to data check on /gfsf2
[2012-03-07 18:31:15.510955] D [afr-lk-common.c:1427:afr_nonblocking_inodelk] 0-datastore-replicate-1: attempting data lock range 0 0 by 48ff7f0cff7f0000
[2012-03-07 18:31:15.512096] D [afr-self-heal-data.c:1130:afr_sh_data_post_nonblocking_inodelk_cbk] 0-datastore-replicate-1: Non Blocking data inodelks done for /gfsf2 by 48ff7f0cff7f0000. Proceeding to self-heal
Should not trigger self-heal
CHANGE: http://review.gluster.com/2992 (glusterd: Disabled self-heal on clear-locks internal mount) merged in master by Anand Avati (email@example.com)
Bug is fixed . verified on 3.3.0qa39