CTDB configuration question

ronnie sahlberg ronniesahlberg at gmail.com
Sun Jun 29 23:30:03 GMT 2008


On Sat, Jun 28, 2008 at 7:17 AM, Collins, Kevin [Beeline]
<KCollins at chevron.com> wrote:
> I have now configured both CIFS and NFS with CTDB and it is working to a
> certain extent:
>
> I have an NFS export node 0 mounted on a client.
>
> On the client, I start a copy from the mount.
>
> I run 'ctdb disable -n 0' from one of the nodes. I use 'ctdb status' and
> 'ctdb ip' to see that node 1 has taken over - my copy continues as
> expected.
>
> I run 'ctdb enable -n 0' from one of the nodes. The copy stops and hangs
> for about 5 minutes before resuming. This obviously seems to be
> unexpected behavior. Both nodes are configured the same.

This happens also if you start by attaching to node1, then you disable
node 1 and it fails over to node 0 and everything is fine?
but when you fail back to node 1 again after enabling it you have an outage?

Can you please send me the output of netstat -a -n on all three hosts
client, node0 and node1 before you disable node0, after you have
disabled node0 and after you have re-enabled node0 again and have the
issue.


Oh,   can you verify that you have IPv6 completely disabled?


>
> In /var/log/messages on both systems (at startup) I was seeing:
>
> rpc.statd[6101]: statd: failed to bind to outgoing port, 596
> falling back on randomly chosen port
>
> I was thinking it might be a privileged port issue since the process is
> running as user "rpcuser" and the port is less than 1024. I changed the
> port to 4596, but I still see:
>
> statd: failed to bind to outgoing port, 4596        falling back on
> randomly chosen port
>
> However, I just noticed that on node 0 I see this message every time
> statd restarts (via the enable), but I only see it on node 1 at the
> initial startup of statd (i.e., 'service ctdb start').
>
> I have confirmed (with lsof) on both nodes that the port is not in use.
>
> Any thoughts?
>
> Thanks,
>
> Kevin
>
> -----Original Message-----
> From: samba-technical-bounces+kcollins=chevron.com at lists.samba.org
> [mailto:samba-technical-bounces+kcollins=chevron.com at lists.samba.org] On
> Behalf Of Collins, Kevin [Beeline]
> Sent: Thursday, June 26, 2008 8:28 AM
> To: ronnie sahlberg
> Cc: samba-technical at lists.samba.org
> Subject: RE: CTDB configuration question
>
> Excellent - thanks!
>
> -----Original Message-----
> From: ronnie sahlberg [mailto:ronniesahlberg at gmail.com]
> Sent: Wednesday, June 25, 2008 7:40 PM
> To: Collins, Kevin [Beeline]
> Cc: samba-technical at lists.samba.org
> Subject: Re: CTDB configuration question
>
> On Thu, Jun 26, 2008 at 1:21 AM, Collins, Kevin [Beeline]
> <KCollins at chevron.com> wrote:
>> Ok, thanks. Do you have any suggestions on the other questions I had?
>>
>> Additionally, if I am using the first doc, then where is
>> STATD_SHARED_DIRECTORY being used? It isn't being passed to the statd
>> command that I can tell...
>
> It is used by
> /etc/ctdb/statd-callout
>


More information about the samba-technical mailing list