This is a discussion on Re: MaxClients and Apache::DBI ? - modperl ; On Aug 8, 2006, at 4:16 AM, Philip M. Gollucci wrote: > That says start 5 servers, I must have least 5 idle and no more > then 10 idle. > You can simultaneously serve 75 clients. which means i ...
On Aug 8, 2006, at 4:16 AM, Philip M. Gollucci wrote:
> That says start 5 servers, I must have least 5 idle and no more
> then 10 idle.
> You can simultaneously serve 75 clients.
which means i can have an 85 server max?
> I Assume you are you're using Apache::Session::Postgres,
> I assume your passing DBH => $dbh, and LockDBH => $dbh,
> Each apache child gets its own DBI connection from connect_on_init
> () if you are using it.
> Well only really 75 of them should be active at once the should
> queue according to your mpm... Which is probably
> prefork, so they just sit in the parent waiting for a child to be
prefork yes. and yes.
> I believe you should be at 75*2 = 150 asumming I read that correctly.
thats what i thought
> If you have Apache(2)::Reload on, you'll blow the Apache:BI handle
no, i'm benching this on my production machine- no reload- which is
odd. so i can't figure out how i'm hitting the postgres max.
I'm going to look into timeout. I'm also going to look into how DBI
is pulling the lock connection-- perhaps it's sending a modified
connect string than the other handles. i didn't think to look there
before you mentioned it.
I spent a few hours rewriting code before i posted, consolidated the
misc other db handles into a single one -- i had to get them all into
autocommit=0, and make sure nothing tries to write off a read only
handle from my factory class.
i'd run the sessions off the same db handle as the rest of my app,
but im uneasy about having that clash with transactional data.