[Gluster-users] Geo-replication big logs and large number of pending files

[Atin Mukherjee]

CCing Geo-rep folks.

~Atin

On 02/24/2015 08:04 PM, wodel youchi wrote:
> Hi,
> I have a 03 nodes setup (Centos7x64 latest updates, glusterfs 3.6.1 latest updates).
> Master: 02 nodes  (g1 and g2) are on replicated mode with 02 volumes data1 and data2, each volume is constituted of one brick.Slave: the 3rd node (g3) is for Geo-Rep with also 2 volumes slavedata1 and slavedata2I am using the geo-rep with a user geoaccount1 and group geogroup1.
> the setup was successfully made and geo-rep started. 
> 
> Problems:- After some days, I've found the geo-rep in a faulty state, the reason /var was full in g1 and g3 the slave node.the ssh log file for geo-replication-slave on g3 was full with (11Go):
> [2015-02-24 11:29:26.526285] W [client-rpc-fops.c:172:client3_3_symlink_cbk] 0-slavedata2-client-0: remote operation failed: File exists. Path: (<gfid:ce5d8b13-1961-4126-93e8-e4ee2fd6b34d>/S15bind9 to ../init.d/bind9)
> [2015-02-24 11:29:26.526297] W [fuse-bridge.c:1261:fuse_err_cbk] 0-glusterfs-fuse: 1100: SETXATTR() /.gfid/ce5d8b13-1961-4126-93e8-e4ee2fd6b34d => -1 (File exists)
> [2015-02-24 11:29:26.526602] W [client-rpc-fops.c:172:client3_3_symlink_cbk] 0-slavedata2-client-0: remote operation failed: File exists. Path: (<gfid:ce5d8b13-1961-4126-93e8-e4ee2fd6b34d>/S20modules_dep.sh to ../init.d/modules_dep.sh)
> [2015-02-24 11:29:26.526618] W [fuse-bridge.c:1261:fuse_err_cbk] 0-glusterfs-fuse: 1101: SETXATTR() /.gfid/ce5d8b13-1961-4126-93e8-e4ee2fd6b34d => -1 (File exists)
> I emptied the log files on both servers, then I modified the logrotate conf file for geo-repl on all nodes from rotate 52 to 
> rotate 7
> size 50M
> Does geo-rep produce such big logs?
> 
> the modifications worked for g1 and g2, but I had a problem with g3[root at glustersrv3 logrotate.d]# logrotate -f /etc/logrotate.d/glusterfs-georep
> error: skipping "/var/log/glusterfs/geo-replication-slaves/967ddac3-af34-4c70-8d2b-eb201ebb645d:gluster%3A%2F%2F127.0.0.1%3Aslavedata1.gluster.log" because parent directory has insecure permissions (It's world writable or writable by group which is not "root") Set "su" directive in config file to tell logrotate which user/group should be used for rotation
> So I added these two lines to the /etc/logrotate.d/glusterfs-georepsu root geogroup1
>  And now it seems working, is that correct?
> After cleaning up the logs, I've tried to restart the geo-rep but didn't succeed: no active session between g1 and g3 erro, so I had to restart the glusterfs daemon on all three nodes.
> After the geo-rep was restarted and the its state became stable, I did a geo-rep status detail and I got this
> [root at glustersrv1 ~]# gluster volume geo-replication data1Â  geoaccount1 at gserver3.domain.tld::slavedata1 status detail
> 
> MASTER NODEÂ Â Â Â Â Â Â Â Â Â Â  MASTER VOLÂ Â Â  MASTER BRICKÂ Â Â Â Â Â Â Â  SLAVEÂ Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  STATUSÂ Â Â Â  CHECKPOINT STATUSÂ Â Â  CRAWL STATUSÂ Â Â  FILES SYNCDÂ Â Â  FILES PENDINGÂ Â Â  BYTES PENDINGÂ Â Â  DELETES PENDINGÂ Â Â  FILES SKIPPED
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> glustersrv1.domain.tld    data1         /mnt/brick1/brick    gserver3.domain.tld::slavedata1    Active     N/A                  Hybrid Crawl    25784          8191             0                0                  0
> glustersrv2.domain.tld    data1         /mnt/brick1/brick    gserver3.domain.tld::slavedata1    Passive    N/A                  N/A             0              0                0                0                  0
> [root at glustersrv1 ~]# gluster volume geo-replication data2Â  geoaccount1 at gserver3.domain.tld::slavedata2 status detail
> 
> MASTER NODEÂ Â Â Â Â Â Â Â Â Â Â  MASTER VOLÂ Â Â  MASTER BRICKÂ Â Â Â Â Â Â Â  SLAVEÂ Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  STATUSÂ Â Â Â  CHECKPOINT STATUSÂ Â Â  CRAWL STATUSÂ Â Â  FILES SYNCDÂ Â Â  FILES PENDINGÂ Â Â  BYTES PENDINGÂ Â Â  DELETES PENDINGÂ Â Â  FILES SKIPPED
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> glustersrv1.domain.tld    data2         /mnt/brick2/brick    gserver3.domain.tld::slavedata2    Active     N/A                  Hybrid Crawl    11768408       8191             0                0                  3833
> glustersrv2.domain.tld    data2         /mnt/brick2/brick    gserver3.domain.tld::slavedata2    Passive    N/A                  N/A             0              0                0                0                  0
> What does it mean  FILES PENDING? because this number didn't change after 1hour from restarting the geo-rep, I thought that it will decrease over time but it didn't.And what does mean FILES SKIPPED?
> I tried another thing, I stopped the geo-rep, stopped the volumes on g3 then deleted them.then I cleaned up the .glusterfs directory on both bricks and deleted all the glusterfs attributes on them with setfattr command, but I did not delete my data (files and directories).
> then I recreated the slave volumes, started them and finally restarted the geo-rep, after the initialization and stabilization I got the same result from status command on geo-rep, the same values on FILES PENDING and FILES SKIPPED
> is that ok? how can I be sure that I have all my data on g3?
> 
> thanks in advance
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
> 

-- 
~Atin