Ticket #101 (Fixed)Fri Dec 15 09:21:20 UTC 2006
Hub falls over with many connections
Reported by: | Andrew Hodgkinson (6) | Severity: | Normal |
Part: | Web site: Hub (single sign-on mechanism) | Release: | 2nd public site release |
Milestone: | 2nd public site release completed | Status | Fixed |
Details by Andrew Hodgkinson (6):
Either it is because of a great many concurrent connections or because the Hub Gem is failing to close files, but Hub fell over during the night due to a “too many open files” error on .rnd.
DRb service to supply the data instead?
As an interim measure, a uschedule services/clean_start call once a night might be useful.
Changelog:
Modified by Andrew Hodgkinson (6) Sat, December 23 2006 - 17:18:20 GMT
- Severity changed from Critical to Normal
Changeset #165 addresses the most serious part of the ticket. Files are no longer left open either permanently or, as was most likely the case, until the next garbage collection sweep- happening too infrequently to prevent “Too many open files” errors.
The ticket severity has been downgraded but the ticket is left open since, as the change log in Changeset #165 mentions, there are efficiency problems with Hub’s use of the random file source. These need sorting out; the DRb server may provide the answer.
Modified by Andrew Hodgkinson (6) Wed, April 04 2007 - 18:27:48 GMT
- Status changed from Open to Fixed
Changeset #192 makes use of the random file considerably more efficient. Changeset #191 implements a similar change to the Hub application.