BUG: Resource issues with OMERO.tables

[spli]

I've been seeing some strange behaviour with OMERO.tables when making a large number of calls to add data from OMERO.scripts. This includes scripts exiting due to a failure to communicate with the Tables service, the Tables service using 100% CPU even after all clients have disconnected, apparent out-of-memory errors with theTables service (manifested as an OS X "Problem Report for Python" dialog).

Further investigation suggets this is due to opening/closing a table too frequently. For instance, reusing a connection/handle to a table instead of opening and closing the table for every operation (read, addData) seems to work fine. Still a bug though.

[spli]

Turns out this (or something similar) is a problem even when the table connections are reused. Occasionally the Tables service is killed and restarts automatically, the only sign is a "Python quit unexpectedly" OS X dialog, with no obvious error log. Tables-0.log only has a ***** ... Waiting 10000 ms on startup indicating the service was restarted with no error prior to this. master.err only has a failed connection from the scripts trying to connect:

01/28/13 16:09:17.917 OMERO.Glacier2: warning: dispatch exception: TcpTransceiver.cpp:158: Ice::ConnectionLostException:
connection lost: recv() returned zero
identity: 7D992E29-C70F-4C25-A3CB-284C8A9824C2
facet: 
operation: getWhereList
01/28/13 16:09:17.919 OMERO.Glacier2: warning: dispatch exception: TcpTransceiver.cpp:158: Ice::ConnectionLostException:
connection lost: recv() returned zero
identity: 986F88C9-E3CD-467B-B2F9-52E04B9324C2
facet: 
operation: getWhereList

[jburel]

Referencing ticket #9944 has changed sprint.

[jburel]

Referencing ticket #9944 has changed sprint.