Slony1 PostgreSQL Replication - Re: sl_log_1 not getting cleared when no replica

Author

Re: sl_log_1 not getting cleared when no replica

Christopher Browne

2006-02-25, 9:50 am

Vivek Khera wrote:

>A couple of weeks ago my replica node got *way* behind for some
>reason. Still haven't figured out what caused it. There were over
>20m entries in sl_log_1 and it was taking on the order of 70 seconds
>to process each event. Since it was never gonna catch up, I decided
>to drop the replica.
>
>So I did a DROP NODE and it cleared out the sl_node and sl_listen,
>etc, tables as expected, and the replica's slon exited.
>
>However, sl_log_1 was never cleared. It just kept on growing and
>growing. It seems not to make sense that any new entries would be
>added.
>
>Ultimately, I did a uninstall node on the master to clear it all out.
>
>Pg 8.0.4 on master, Slony 1.1.5, FreeBSD 5.4.
>
>
>
I guess it's never seeing an acknowledgement from anyone else, so events
are being regarded as "not confirmed."

That's probably worthy of some sort of extra code path in the cleanup
loop; if there are no nodes other than the local one, old events should
be treated as confirmed...

That needs a bit of thought...

Jan Wieck

2006-02-25, 9:50 am

On 2/23/2006 9:33 AM, Vivek Khera wrote:

> On Feb 22, 2006, at 6:06 PM, Christopher Browne wrote:
>
>
> should i file a formal bug report?

Yes, please do so. In the meantime, a workaround would have been to
recreate the replica and start its slon daemon, but NOT yet subscribe
it. This would have caused events to be confirmed and the cleanup
eventually to take place.

Jan

--
#===================
====================
====================
===========#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#===================
====================
=========== JanWieck- bwPqjjyvM7QAvxtiuMwx
3w@public.gmane.org #

Sponsored Links