Gluster users - Jul 2013 - Healing in glusterfs 3.3.1

Hi,

I have a glusterfs 3.3.1 setup with 2 servers and a replicated volume used
by 4 clients.

Sometimes from some clients I can't access some of the files. After I force
a full heal on the brick I see several files healed. Is this behavior
normal?

Thanks
-- 
Paulo Silva <paulojjs at gmail.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130724/f367eb48/attachment.html>

On 07/24/2013 07:27 PM, Paulo Silva wrote:
> Hi,
>
> I have a glusterfs 3.3.1 setup with 2 servers and a replicated volume
> used by 4 clients.
>
> Sometimes from some clients I can't access some of the files. After I
> force a full heal on the brick I see several files healed. Is this
> behavior normal?
>
With 3.3.1, healing happens in two modes:

a) index heal - based on changes recorded in indices on bricks. The 
indices are consumed by self-heal daemons on each server for performing 
healing and is the default mode of operation.

b) full heal - works through a full crawl of the volume. In this mode, 
self-heal daemon inspects changelogs of all files and directories it 
encounters during a crawl. This mode is triggered when volume heal full 
command is used.

Normally, a) is faster and addresses most healing cases. Do you have 
more details on files that require a full heal to be performed for 
healing completion?

Regards,
Vijay

Hi, I have the same issue. Sometimes, the files are not getting replicated
between the bricks.
Glusterd service and glusterfsd are running. When the files replication is not
happening, these are the outcomes of the commands.

Service glusterfsd  does not restart. It just stops when I try to restart it.
Also self-heal commands says no heal required. I have to reboot the server for
the file replication to start off.

Root> ps -ef | grep glusterfsd (gives me the following services running)

root      1928     1  0 Aug12 ?        00:03:00 /usr/sbin/glusterfsd -s
localhost --volfile-id gfscloud.KWTTESTGSNODE002.mnt-cloud_gfs -p
/var/lib/glusterd/vols/gfscloud/run/KWTTESTGSNODE002-mnt-cloud_gfs.pid -S
/tmp/216f54af381565662c717863ab3b6312.socket --brick-name /mnt/cloud_gfs -l
/var/log/glusterfs/bricks/mnt-cloud_gfs.log --xlator-option
*-posix.glusterd-uuid=684855a9-5208-4480-a52b-f2bc8caeb992 --brick-port 24011
--xlator-option gfscloud-server.listen-port=24011

Root> gluster volume status.

Status of volume: gfscloud
Gluster process                                         Port    Online  Pid
------------------------------------------------------------------------------
Brick KWTTESTGSNODE002:/mnt/cloud_gfs                   24011   Y       1928
Brick ZAJILTESTGSNODE001:/mnt/cloud_gfs                 24011   Y       1588
NFS Server on localhost                                 38467   Y       1934
Self-heal Daemon on localhost                           N/A     Y       1940
NFS Server on ZAJILTESTGSNODE001                        38467   Y       1611
Self-heal Daemon on ZAJILTESTGSNODE001                  N/A     Y       1620


Thanks & Regards,
Bobby Jacob

-----Original Message-----
From: gluster-users-bounces at gluster.org [mailto:gluster-users-bounces at
gluster.org] On Behalf Of Vijay Bellur
Sent: Saturday, July 27, 2013 6:35 PM
To: Paulo Silva
Cc: gluster-users at gluster.org
Subject: Re: [Gluster-users] Healing in glusterfs 3.3.1

On 07/24/2013 07:27 PM, Paulo Silva wrote:
> Hi,
>
> I have a glusterfs 3.3.1 setup with 2 servers and a replicated volume 
> used by 4 clients.
>
> Sometimes from some clients I can't access some of the files. After I 
> force a full heal on the brick I see several files healed. Is this 
> behavior normal?
>
With 3.3.1, healing happens in two modes:

a) index heal - based on changes recorded in indices on bricks. The indices are
consumed by self-heal daemons on each server for performing healing and is the
default mode of operation.

b) full heal - works through a full crawl of the volume. In this mode, self-heal
daemon inspects changelogs of all files and directories it encounters during a
crawl. This mode is triggered when volume heal full command is used.

Normally, a) is faster and addresses most healing cases. Do you have more
details on files that require a full heal to be performed for healing
completion?

Regards,
Vijay



_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://supercolony.gluster.org/mailman/listinfo/gluster-users