| Title | A hanging query can lock up the defect tracker |
| Status | suspended |
| Priority | essential |
| Assigned user | Nick Barnes |
| Organization | Ravenbrook |
| Description | The replicator may take some locks in the defect tracker during a replication poll. If the replicator hangs during the poll (for instance because a query made to Perforce or to the defect tracker hangs), these locks may interfere with normal operation of the defect tracker. This applies in particular to Bugzilla, where the replicator locks most of the Bugzilla tables in order to ensure it gets an atomic view of the Bugzilla state. A customer having problems with Perforce (caused by a Linux kernel bug) found that a Perforce query made by the replicator was hanging and that this prevented normal operation of Bugzilla [1]. |
| Analysis | All queries made by the replicator should have some sort of time out. Hitting the time out should cause a poll failure. |
| How found | customer |
| Evidence | [1] <http://info.ravenbrook.com/mail/2003/05/19/20-59-59/0.txt> |
| Observed in | 1.5.3 |
| Created by | Nick Barnes |
| Created on | 2003-05-20 17:14:00 |
| Last modified by | Nick Barnes |
| Last modified on | 2018-07-05 17:28:07 |
| History | 2003-05-20 NB Created. 2018-07-05 NB Suspended because the P4DTI is obsolete. |