On Mon, Nov 1, 2010 at 11:53 AM, Christopher Browne
<cbbrowne at ca.afilias.info> wrote:
> If the slon processes are running in UTC/GMT, as is recommended, then
> there's no issue whatever, because UTC/GMT do not adjust for DST.
>

I disagree.  At least at some point in history this happened to me at
the DST switchover:

[rt]% psql -U postgres -x -c 'select * from _rt.sl_status' rt3
-[ RECORD 1 ]-------------+---------------------------
st_origin                 | 20
st_received               | 10
st_last_event             | 4896924
st_last_event_ts          | 2007-11-04 01:08:57.010633
st_last_received          | 4896923
st_last_received_ts       | 2007-11-04 01:08:53.002901
st_last_received_event_ts | 2007-11-04 01:08:53.002292
st_lag_num_events         | 1
st_lag_time               | -00:59:54.583198

[rt]% date
Sun Nov  4 01:09:01 EDT 2007


I'm not 100% convinced that having the slon run in UTC is sufficient
for there to be "no problems" as the above is totally based on the
server's idea of the time.

For that full hour the lag time was off by one hour.  It made for some
fun nagios alerts.

I guess this sunday night we can try again and see what happens... :)
I don't recall what slony version that the above was from.  Whatever
was current at that date.  And it would have been on Pg 8.1.