Description of problem: Rename of nfs mounted file failed due to unknown rename retries at GlusterFS. The successive ‘rename’ requests expect first one returned with error cause ‘acquiring inodelk failed Resource temporarily unavailable’.( mv: cannot move `/opt/swd/file1' to `/opt/swd/file2': Remote I/O error) How reproducible: Rename a nfs file continually, issue could be reproducible some times. Ex: rename file1 to file2 and file2 to file1 continually. Steps to Reproduce: 1)From client where gluster-nfs mount exist; create a temporary file (say file1). touch file1.txt 2)Execute the below script from client where gluster-nfs mount exist, which renames the file1 to file2 and file2 to file1 continually. Script: #!/bin/bash while [ 1 ] do mv /opt/swd/file1 /opt/swd/file2 #echo $? mv /opt/swd/file2 /opt/swd/file1 #echo $? done 3)Issue could be reproducible after some time. Actual results: [2017-03-15 17:07:28.810878] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file1 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file2 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:28.830561] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file2 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file1 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:28.847692] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file1 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file2 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:28.865172] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file2 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file1 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:29.087671] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file1 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file2 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:29.087932] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file1 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file2 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:29.088570] W [dht-rename.c:1144:dht_rename_lock_cbk] 0-_opt_swd-dht: acquiring inodelk failed (Resource temporarily unavailable) rename (/file1:3cbb4683-04f9-4383-9330-b8285f668dc1:_opt_swd-replicate-0 /file2:00000000-0000-0000-0000-000000000000:(null)), returning EBUSY [2017-03-15 17:07:29.088615] W [nfs3.c:3673:nfs3svc_rename_cbk] 0-nfs: 8207de87: rename /file1 -> /file2 => -1 (Device or resource busy) [2017-03-15 17:07:29.088637] W [nfs3-helpers.c:3401:nfs3_log_common_res] 0-nfs-nfsv3: XID: 8207de87, RENAME: NFS: 10006(Error occurred on the server or IO Error), POSIX: 14(Bad address) [2017-03-15 17:07:29.107799] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file2 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file1 (hash=_opt_swd-replicate-0/cache=<nul>) [2017-03-15 17:07:29.125460] I [dht-rename.c:1316:dht_rename] 0-_opt_swd-dht: renaming /file1 (hash=_opt_swd-replicate-0/cache=_opt_swd-replicate-0) => /file2 (hash=_opt_swd-replicate-0/cache=<nul>) Expected results: Rename of nfs file shouldn’t fail due to immediate unknown rename retries in GlusterFS.
Hi, 3.6.0 has reached the end of life. Please feel free to open it on the current versions if you come across the bug. I'm clsong this bug as it has reached EOL.
Hi, 3.6.0 has reached the end of life. Please feel free to open it on the current versions if you come across the bug. I'm closing this bug as it has reached EOL.