[Skytools-users] How to correctly temporarily disable and resume replication in londiste 2.1

[Marko Kreen]

On 7/28/09, Mohanaraj Gopala Krishnan <mohangk at gmail.com> wrote:
> Sorry if this is a repeat question (my googling pulled nothing up).
>
> 1.What is the recommended way to temporary disable replication? During the
> period that the replication is disabled the master would continue to get
> updated however. I would also want the replication to resume upon re
> enabling the replication
>
> Do I just do a
>
> londiste.py p-to-s.ini --stop
> pgqadm.py ticker.ini --stop
> (I assume when shutting down we should bring the londiste daemon down first
> ? ) and then to resume replication I do a
>
> pgqadm.py ticker.ini ticker -d
> londiste.py p-to-s.ini replay -d
>
> Is there anything I would need to do before or right after restarting the
> daemons to ensure that the replication resumes correctly?

Stopping Londiste is enough.  Never stop ticker for a long period - the
queue will feel rather sick afterwards, all events that happened in the
meantime will end up in single batch.

> 2. If the replication is left disabled (and the master database keeps
> getting updated with data during this period ) for a long period will that
> result in any issue ?

If ticker is running then there should be no problems.

> 3. In the londiste tutorial on the postgresql wiki
> (http://wiki.postgresql.org/wiki/Londiste_Tutorial#Londiste_is_eating_all_my_CPU_and_lag_is_raising)
> there is a mention of "This happens for example if you had a nightly glitch
> and the on-call sysadm forgot to restart the ticker." This seems to mean to
> me that leaving the ticker off might create an issue. But is this a case
> where only the ticker was off but the londiste daemon was still on?

If ticker is stopped then Londiste will get no batches thus it also
is effectively stopped.

-- 
marko