postgres at dac.e4ward.com postgres at dac.e4ward.com
Tue Dec 4 13:20:28 PST 2007
On Dec 4, 2007 7:05 PM,  <slony1-general at lists.slony.info> wrote:
> <postgres at dac.e4ward.com> writes:
> > I've correctly dropped my node2 (still have node1 which is the set1
> > origin and node3 which is a slave) using this command:
> >
> > slonik_drop_node node2 | slonik
> >
> > The documentation says that this would try to kill the running slon
> > for that node but it didn't. Could it be because postgres is running
> > with user postgres and the slons are running as root?
> >
> > It correctly dropped the node because it's not listed anymore on the
> > origins sl_node, sl_subscribe ad sl_path tables. But it's still is
> > listed when I ask for replication status on sl_status:
> >
> > # SELECT * from _replication.sl_status ;
> >  st_origin | st_received | st_last_event |      st_last_event_ts
> > | st_last_received |    st_last_received_ts     |
> > st_last_received_event_ts  | st_lag_num_events |   st_lag_time
> > -----------+-------------+---------------+----------------------------+------------------+----------------------------+----------------------------+-------------------+-----------------
> >          1 |           2 |         34498 | 2007-12-04 13:36:29.145529
> > |            34422 | 2007-12-04 13:23:46.859689 | 2007-12-04
> > 13:23:46.835432 |                76 | 00:12:48.167806
> >          1 |           3 |         34498 | 2007-12-04 13:36:29.145529
> > |            34497 | 2007-12-04 13:36:23.466897 | 2007-12-04
> > 13:36:19.108237 |                 1 | 00:00:15.895001
> > (2 rows)
> >
> > Is this going to be a problem if I create a new node and assign it
> > number 2? Will slony think it's the old node2 coming back again and
> > just try to sync those 76 unsynced events?
>
> The dead node will go away from that view in ~10 minutes, the next
> time the cleanup thread runs against the node where you ran that
> query.
>
> You will be able to create node #2 again, but I'd want to wait about
> 1/2h in order for traces of the "old instantiation" to get cleared out
> on all the nodes.  If you recreated node 2 right away, I could see
> there being a fair bit of confusion.

Cool answer. Indeed it disappeared automatically after a few minutes.

I recreated the node after that and all went smoothly.

Thanks.

-- 
Diego Algorta Casamayou
http://www.oboxodo.com - http://diego.algorta.net


More information about the Slony1-general mailing list