[Gluster-users] How do I slow down client reconnect attempts?

Amar Tumballi atumball at redhat.com
Sun May 21 03:33:15 UTC 2017


On Sat, 20 May 2017 at 1:03 AM, Jim Goven <radu.gluster at ohmi.org> wrote:

> Hello,
>
> I would like to use glusterfs to replicate a volume between my desktop
> and my laptop, so that when I am away from home (and likely
> disconnected), I can still access some project files. The desktop is on
> most of the time, but the laptop is rarely on.
>
> I have successfully set up gluster to accomplish this goal, but I am
> looking to improve the disconnected behaviour of the desktop.
>
> While the laptop is disconnected (ie, most of the time), the desktop
> continuously re-tries to connect to the laptop, but in doing so it
> generates too much DNS traffic trying to find the laptop - it looks like
> there is no retry delay at all. In turn, the gluster client log files
> grow very fast, the DNS resolver logs also grow fast, and there is a lot
> of unnecessary network chatter.
>
> I would like to add a delay of between subsequent re-attempts to
> reconnect, but I can't find any such option at
> http://gluster.readthedocs.io/...Managing%20Volumes
>
> The question is: Is there a way to slow down gluster's attempts to
> reconnect?
>
> I looked through the source as well, and I can't find a delay mechanism.
>
> Ideally the reconnect attempts would follow an exponential back-off
> schedule with a configurable maximum delay, but a fixed, configurable
> delay would work too.
>
> In case a developer reads this, I would be inclined in implementing
> exponential backoff myself, and submitting a patch, but would appreciate
> a brief advice of where in the codebase this backoff should be inserted.



Thanks for the offer. Check https://github.com/gluster/glusterfs/issues/152

Currently this is fixed number, you can either make it a xlator option or
exponential backoff.

If you make it as xlator option, this should be in protocol/client
codebase, and then handle it in rpc_clnt.c code.

Regards,
Amar

>
> Thank you
> jg
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://lists.gluster.org/mailman/listinfo/gluster-users
>
-- 
Amar Tumballi (amarts)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170521/2a880ab2/attachment.html>


More information about the Gluster-users mailing list