[QFJ-571] QFJ Timer - JDBCStore reset issues Created: 20/Dec/10 Updated: 08/Apr/14 Resolved: 08/Apr/14 |
|
Status: | Closed |
Project: | QuickFIX/J |
Component/s: | Engine |
Affects Version/s: | 1.5.0 |
Fix Version/s: | None |
Type: | Bug | Priority: | Major |
Reporter: | Vid Cheruvu | Assignee: | Unassigned |
Resolution: | Duplicate | Votes: | 0 |
Labels: | None |
Issue Links: |
|
Description |
Hi, I have configured QuickFixJ to use JDBC store. In doing so, we have seen an issue with Proxool where it kills active thread that are longer than 5 minutes. We have no way to change the proxool's maximum-active-time property in Quickfixj configuration. When proxool kills the thread, QFJ failed to perform Scheduled Session reset. Please see the error log as shown below. 2010-12-18 09:32:07,312 WARN [HouseKeeper] HouseKeeper.sweep:149 - #0007 was active for 5559 milliseconds and has been removed automaticaly. Th Kind Regards, |
Comments |
Comment by Christoph John [ 16/Oct/12 ] |
There were some synchronization problems around the Session.reset() method which might have caused this. Could you please try with QF/J 1.5.3 as soon as it is released (October/November 2012) and let us know if the problem still persists? Thanks, Chris. |
Comment by Lucian Hada [ 12/Nov/12 ] |
Hi, we have the same issue using 1.5.0 and I don't see 1.5.3 out yet. Any chance this will get away with 1.5.2? |
Comment by Christoph John [ 12/Nov/12 ] |
It is likely that the problem is still present on 1.5.2. I'm not even sure that it is fixed with 1.5.3 but it is quite likely since |
Comment by Lucian Hada [ 12/Nov/12 ] |
Thanks for answer. Are there any recommended workarounds? I didn't managed to reproduce this in a controlled test but it seems to me that once it occurred the sequence won't reset even if we restart the application and we don't see the exception anymore. Using MySQL btw. |
Comment by Christoph John [ 13/Nov/12 ] |
In the other issues I mentioned the problem was that there were messages sent and the session reset was done at the same time. Maybe you could define with your counterparty that they log off some time before your session's EndTime is reached. That way you could make sure that there are no messages exchanged while QF/J tries to reset the seqnums. This is the only workaround that I could think of at the moment. |