| Title | Exponential backoff looks like failure |
| Status | suspended |
| Priority | optional |
| Assigned user | Nick Barnes |
| Organization | Ravenbrook |
| Description | When a replicator poll fails, the replicator "backs off", doubling the period between polls. This is to prevent an e-mail deluge (job000215). If polls continue to fail, the poll period gets very long, and once it gets more than a day or so the replicator appears to have stopped. There isn't any indication of this back-off in the poll failure message, and an administrator who forgets the way the exponential back-off works might well think that the replicator had stalled entirely [1]. This was, in fact, job000482. |
| Analysis | 1. Stop doubling once the poll period gets to a day or so. 2. State, near the top of the poll failure message, the new poll period and expected next polling time. 3. Also put this information in the log, at a level higher than DEBUG, as the last log item before sleeping. |
| How found | customer |
| Evidence | [1] <http://info.ravenbrook.com/mail/2002/03/27/13-36-48/0.txt> |
| Observed in | 1.3.3 |
| Introduced in | 1.0.1 |
| Created by | Nick Barnes |
| Created on | 2002-03-27 15:50:08 |
| Last modified by | Nick Barnes |
| Last modified on | 2018-07-05 17:27:58 |
| History | 2002-03-27 NB Created. 2018-07-05 NB Suspended because the P4DTI is obsolete. |