[egenix-users] mxODBC issues with MS SQLServer, blocking question and ZODB bloat issue

Ivo van der Wijk vladdrac at gmail.com
Thu Sep 14 17:14:11 CEST 2006


Hi All,

I've been looking into an issue the last few days at a customer, where
a Zope/Plone instance is suffering from occasional 'hanging'. The
system uses mxODBC / mxODBCDA / ZSQL extensively, and the problem
'feels' alot asif all of the threads end up blocking on a database
connection. When the system hangs, there is hardly any CPU usage, and
no threads are available to handle new requests (i.e. ZMI management
access). The queries that are used are trivial.

We've already tried the threading fix as described at
https://www.egenix.com/mailman-archives/egenix-users/2005-October/113974.html,
but that didn't seem to fix the problem. Currently, IODBC in stead of
unixODBC is being tested.

The current configuration is as follows:

mxODBC 1.0.9 (commercial license)
Zope 2.8.5
Plone 2.1.2
unixODBC
freesoft OOB

if anyone has similar experiences (I saw a post recently by robert
rotterman to the list describing a similar issue) or any ideas on how
to test / trace / fix this issue, please let me know :)

Additionally, I would like to know if there is any way to have a
timeout set on the max. duration of a query through mxODBC.

Lastly, while debugging issues and the software in general, we noticed
a small, mysterious ZODB bloat. It appears that when the mxODBCDA has
'connect on demand' enabled, it will occassionally create new
transactions, possibly because of some internal, persisted,
bookkeeping of the current connection state. You might want to look
into this, having read only queries / pageviews result in ZODB writes
is usually a bad thing IMHO.

Regards,

 Ivo van der Wijk


-- 
Drs. I.R. van der Wijk / m3r Consultancy B.V.
Linux/Python/Zope/Plone and Open Source solutions
PO-box 51091, 1007 EB Amsterdam, The Netherlands
Email: ivo <at> m3r.nl



More information about the egenix-users mailing list