Hello,
we are using Gluster 3.2.1, distributed setup with two bricks:
Volume Name: s
Type: Distribute
Status: Started
Number of Bricks: 2
Transport-type: tcp
Bricks:
Brick1: gluster01:/data
Brick2: gluster02:/data
Options Reconfigured:
performance.io-thread-count: 8
performance.write-behind-window-size: 4MB
performance.stat-prefetch: off
Partitions /data are using XFS.
Clients use FUSE to mount it. On the client side when writing files to
gluster in logs we see a lot of following errors:
[2011-07-07 13:19:42.713369] W
[dht-common.c:941:dht_lookup_linkfile_cbk] 0-s-dht: /file1: gfid
different on data file on s-client-1
[2011-07-07 13:19:42.713813] I
[dht-common.c:838:dht_lookup_everywhere_cbk] 0-s-dht: deleting stale
linkfile /file1 on s-client-0
[2011-07-07 13:19:42.713874] W
[dht-common.c:781:dht_lookup_everywhere_cbk] 0-s-dht: /file1: gfid
differs on subvolume s-client-1
One brick contains empty sticky file and the other one the correct one:
--------- gluster01---------
-rw-r--r-- 1 suser users 753981 Feb 9 19:24 /file1
--------- gluster02---------
---------T 1 suser users 0 Jul 7 13:40 /file1
When accessing the file through FUSE correct version of file is served.
Could someone clarify why is this happening and how to fix it?
Thanks
--
Emir Imamagic
www.srce.hr