[Bucardo-general] Bucardo design question

[Michelle Sullivan]

Greg,

I'm really running into problems getting this to catch up - the time to
gather the data for the sync is increasing to unworkable levels.  The
problems are not due to Bucardo but my knowledge about the cleanup crons
, and a couple of hardware issues.  These have been solved and I need
for this to catchup really quick.  Can you explain (very quickly) how
Bucardo knows about transactions....

Specifically I have writers in my app that will 'BEGIN' then insert or
update several tables before 'COMMIT'ing...  Does Bucardo see the update
of all the tables as a single transaction?

Does Bucardo know about the foreign keys so that it only inserts rows in
the correct order?

How does Bucardo choose what to sync from each table, point in time (ie
now) or to the last transaction?  I'm happy to trawl the code if you
don't have time, but would it be possible to limit the number of
transactions in the sync (of course my previous questions could answer
that question for me ;-) ).... I just need to know whether I'd be
wasting my time as I have 4 days to get this system into production -
which means bucardo has to catchup.  I have already found (and solved)
the reason it got out of sync in the first place, but it's taking around
1-2 days to start writing transactions to the other side of the swap
sync due to the huge number and the fact the sync process uses all
available ram and an extra 16G of swap...! ;-)  I suspect limiting
(temporarily) to updating 100k transactions at a time would speed this
process up immensely.


Thanks in advance,

Michelle