Gluster users - May 2015 - fault tolerance

Hi people,
Now I am using the version of gluster 3.6.2 and I want configure the system for
fault tolerance. The point is that I want have two server in replication mode
and if one server down the client do not note the fault. How I need import the
system in the client for this purpose.

Dear Carlos,

Have a look at http://bit.ly/gadmstp (section "Mounting options"); you
can use the "backupvolfile-server" option for this.

Kind regards, Markus

Am 16.05.2015 um 22:17 schrieb Carlos J. Herrera:
> Hi people,
> Now I am using the version of gluster 3.6.2 and I want configure the
> system for fault tolerance. The point is that I want have two server
> in replication mode and if one server down the client do not note the
> fault. How I need import the system in the client for this purpose.

(@Carlos: Your reply to my answer seemingly did not reach this list, so
I just quote it below and reply to your first post)

Dear Carlos,

According to "man mount.glusterfs", you can provide a colon separated
list of backup servers using the "backup-volfile-servers" option (note
the slightly changed naming):
# mount -t glusterfs
-obackup-volfile-servers=<server2>:...:<serverN>\
    <server1>:/<volname> <mount_point>

This should be equivalent to the following /etc/fstab entry you'd use:
<server1>:/<volname> <mount_point> glusterfs defaults,_netdev,
\
    backup-volfile-servers=<server2>:<server3> 0 0

However, when I tried the above using /usr/sbin/glusterfs version 3.6.3
which maps this to the following call, I saw in the log that the
secondary/third servers were not contacted when the first was
unavailable (i.e., mounting failed in that case):
# /usr/sbin/glusterfs --volfile-server=<server1> \
    --volfile-server=<server2> --volfile-server=<server3> \
    --volfile-id=/<volname> <mount_point>

So while the above answers your question, you might be affected by bug
https://bugzilla.redhat.com/show_bug.cgi?id=1222678 as well.

Kind regards, Markus

Am 18.05.2015 um 02:08 schrieb Carlos J. Herrera:
> Hello people.
> How I can use backupvolfile-server with 3 server??

For the client's not to notice the fault you would need to have three
servers in replica 3, otherwise the filesystem will become read-only with
one node down.

On 17 May 2015 at 03:25, Markus Ueberall <Markus.Ueberall at gmail.com>
wrote:
> Dear Carlos,
>
> Have a look at http://bit.ly/gadmstp (section "Mounting
options"); you
> can use the "backupvolfile-server" option for this.
>
> Kind regards, Markus
>
> Am 16.05.2015 um 22:17 schrieb Carlos J. Herrera:
> > Hi people,
> > Now I am using the version of gluster 3.6.2 and I want configure the
> > system for fault tolerance. The point is that I want have two server
> > in replication mode and if one server down the client do not note the
> > fault. How I need import the system in the client for this purpose.
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150521/e888235c/attachment.html>

On 05/16/2015 01:17 PM, Carlos J. Herrera wrote:
> Hi people,
> Now I am using the version of gluster 3.6.2 and I want configure the system
for fault tolerance. The point is that I want have two server in replication
mode and if one server down the client do not note the fault. How I need import
the system in the client for this purpose.
>
When using a fuse mount, or libgfapi, with a replica volume, fault 
tolerance is built in. Once mounted, the client connects to every brick 
in the volume. The only possible issue may be the ping-timeout default 
of 42 seconds. This is long to give your tcp connection time to recover 
in the event of a temporary network fault. If your server completely 
dies, however, the client will have to wait for this timeout. I 
recommend leaving the timeout long and instead using servers that don't 
die frequently.

The tip the other guys mentioned have to do with mounting the volume if 
the first server in the volume is unavailable at the time of mount. 
Besides backupvolfile-server you can use a round-robin dns entry to 
resolve mounting if a server is down.