[nycbug-talk] Re: MySQL

[Jonathan Vanasco]

On Aug 21, 2005, at 3:18 PM, Hans Zaunere wrote:
> Postfix generally generates a lot of the same queries - many small fast
> queries.  Try enabling query cache.  For simple SELECT statements, you
> should have no problem handling thousands of concurrent connections.

There might also be a query cache or table in postfix itself that you 
can configure to have a longer timeout.
I believe exim does this, but i could be wrong.

On Aug 21, 2005, at 3:22 PM, Francisco Reyes wrote:
> I already installed Mytop in the worst offender. 150+ connections.
> Is there any way to see what each is doing?

That doesn't sound unreasonable or slow.  I think the default 
max_connects for mysql is 150 or 200, and most machines should handle 
that fine
But bob is right about connection pooling.

What's the volume of email that's going through the system?  also, you 
said primarily for postfxi -- does that mean that imap and other mail 
related services aren't working against this db either?

On Aug 21, 2005, at 2:07 PM, Bob Ippolito wrote:
> For whatever reason, this made a measurable difference for us.. but 
> this was a TEXT column, VARCHAR is probably not so bad.  Other than 
> that, our queries use indexed integers to find rows.

if you want speed and disk size is not an issue, abandon VARCHAR  -- 
but only if you can do it on the entire row
tables entirely of  fixed-length records (such as char, int, set) take 
advantage of an internal mysql optimization that has the record 
location computed based on the index data.  once a variable length 
record is in there, it has to 'figure stuff out'.  at least that was 
the case in mysql 2 years ago - but i doubt that changed, as its a 
pretty logical cause and effect.