[Twisted-Python] Making ConnectionPool not pool
hs at ox.cx
Tue Nov 27 07:49:11 EST 2012
> I have a really bad time with the combination of a low-volume service and adbapi.ConnectionPool, pyodbc, FreeTDS and Sybase.
> Basically my connections just time out and fail in weird, generic ways like:
> Error: ('01000', ' [FreeTDS][SQL Server]Unexpected EOF from the server (20017) (SQLEndTran)')
> (but in many others too, there is no real pattern)
> Have you tried enabling reconnects (cp_reconnect=True)? If "select 1" doesn't work with your database you may also have to pass in custom cp_good_sql.
Yep, and the connections *do* heal after dying. I don't have to restart twistd, it just takes a few failed requests to warm up again.
I tried to do a for loop like:
succ = False
for _ in range(self.dbpool.max + 1):
succ = True
if not succ:
raise pyodbc.Error(b'Found no healthy connection.')
…for new connections, but that helped only a bit. I presume once it finds a working connection, it keeps using it. But I still had failure within the session later.
> In my desperation, I’m employing for-loops for the SQL queries now. :(
> Since there isn’t much traffic (yet) I would like to just make ConnectionPool close the connections and re-open fresh ones, as soon as they are necessary.
> Is there some straight-forward way to do that? Or any better approach I’ve overlooked?
> Don't use ConnectionPool at all. Just have a function that does the SQL connect etc usually normal DB-API, and call it with twisted.internet.threads.deferToThread:
> def dbTxn(x):
> conn = db.connect(...)
> cursor = conn.cursor()
> result = cursor.fetchall()
> return result
> deferredResult = deferToThread(dbtxn, argForX)
Sounds like a solid backup plan, thanks!
The only reason I’d prefer to stay with pooling is that I very much expect the traffic to rise and wouldn’t really want to change away and go back later. :(
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Twisted-Python