[Pgpool-general] General pgpool locking issues

[John Black]

We have been testing with pgpool for a little while here and have run into
various locking (or lack there of) issues.

Our setup is 2 databases (pg v8.1.4) on seperate machines with all db
connections connected via pgpool.  We have run into situations where we the
1st or 2nd database was not getting the update or insert fast enough, and a
select query hitting pgpool at the same time causing the a "mismatch" error
causing pgpool to degrade and stop replicating.

Although I assume this must be quite a normal issue, I do not see many
people talking about it here.

Can someone give me a basic rundown of the "best practices" for pg locking
on a pretty database intensive web app?

On Inserts the best practice is to lock the entire table (LOCK TABLE ...) ?

On Updates the best practice is to do a row lock (SELECT ... FOR UPDATE  or
FOR SHARE) ?
Should you do some sort of lock on SELECTS?
Do you recommend everything be wrapped in Transactions?

Any advice is appreciated.

Thank you.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pgfoundry.org/pipermail/pgpool-general/attachments/20060730/866ff7a1/attachment.html