Gluster users - Mar 2014 - GlusterFS cluster of 2 nodes is disconnected after nodes reboot

Hello!

I have 2 nodes with GlusterFS 3.4.2. I created one replica volume using 2
bricks and enabled glusterd autostarts. Also firewall is configured and I
have to run "iptables -F" on nodes after reboot. It is clear that
firewall
should be disabled in LAN, but I'm interested in my case.

Problem: When I reboot both nodes and run "iptables -F" peer status is
still disconnected. I wonder why. After "service glusterd restart"
peer
status is connected. But I have to run "gluster volume heal
<volume-name>"
to make both servers consistent and be able to replicate files. Is there
any way to eliminate this problem?

I read about server-quorum, but it needs 3 or more nodes. Am I right?

Best Regards,
Artem Konvalyuk
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140325/737068d0/attachment.html>

Let's go with the data collection first.

What linux distro ?

Anything special about your network configuration ?

Any chance your server is taking too long to release networking and gluster
is starting before network is ready ?

Can you completely disable iptables and test again ?

I am afraid quorum will not help you if you cannot get this issue
corrected.




On Tue, Mar 25, 2014 at 3:14 PM, ????? ???????? <artret at gmail.com>
wrote:
> Hello!
>
> I have 2 nodes with GlusterFS 3.4.2. I created one replica volume using 2
> bricks and enabled glusterd autostarts. Also firewall is configured and I
> have to run "iptables -F" on nodes after reboot. It is clear that
firewall
> should be disabled in LAN, but I'm interested in my case.
>
> Problem: When I reboot both nodes and run "iptables -F" peer
status is
> still disconnected. I wonder why. After "service glusterd
restart" peer
> status is connected. But I have to run "gluster volume heal
<volume-name>"
> to make both servers consistent and be able to replicate files. Is there
> any way to eliminate this problem?
>
> I read about server-quorum, but it needs 3 or more nodes. Am I right?
>
> Best Regards,
> Artem Konvalyuk
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140325/0b336c35/attachment.html>