Andrew Hammond andrew.george.hammond at gmail.com
Mon Jun 4 17:03:21 PDT 2007
I haven't seen any patch to address this. It seems like a pretty
unlikely corner case and not one that causes data loss. As you note,
there are (at least) two workarounds; either restart the database
(ugly, but probably something that a newbie would try), or manually
delete the entry from sl_nodelock (the "trust me, I know what I'm
doing" approach). So, while it'd be nice to fix it, but I don't think
it's terribly high priority.

If you're keen enough to do the work, a patch would be welcome. :)

A


On 6/4/07, David Rees <drees76 at gmail.com> wrote:
> I saw that there was some talk of 1.2.10 coming real soon now, has
> anyone looked at this issue which I posted about a couple weeks ago?
>
> Let me know if I can provide any more information.
>
> -Dave
>
> On 5/16/07, David Rees <drees76 at gmail.com> wrote:
> > I upgraded my dev setup from pg 8.2.3 to pg 8.2.4 and slony 1.2.8 to
> > 1.2.9 today and it went smoothly, except that one slony node failed to
> > come up with the messages:
> >
> > FATAL  localListenThread: "select "_rep1".cleanupNodelock(); insert
> > into "_rep1".sl_nodelock values (    1, 0, "pg_catalog".
> > pg_backend_pid()); " - ERROR:  duplicate key violates unique
> > constraint "sl_nodelock-pkey"
> > FATAL  Do you already have a slon running against this node?
> > FATAL  Or perhaps a residual idle backend connection from a dead slon?
> >
> > I checked the process list and found no extra slon daemons or residual
> > backend connections, so then I took a look at the sl_nodelock table
> > for the affected node and noticed that the nl_backendpid listed for
> > that node happened to be the same as the pid for the currently running
> > pg daemon itself so the cleanupnodelock function wasn't cleaning up
> > the entry.
> >
> > After restarting postgres the affected slon node came up normally. I
> > suspect I could have manually cleaned out the sl_nodelock table as
> > well.
> >
> > I don't know enough about the _Slony_I_killBackend function, but
> > perhaps it could be improved to detect this situation.
> >
> > -Dave
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general
>


More information about the Slony1-general mailing list