| Summary: | ASTERISK-06849: [post 1.4] SIP Notify with MWI re-sent for all sip users on asterisk restart or reload | ||
| Reporter: | bill (bnwcd) | Labels: | |
| Date Opened: | 2006-04-26 10:53:23 | Date Closed: | 2007-01-23 02:53:51.000-0600 |
| Priority: | Major | Regression? | No |
| Status: | Closed/Complete | Components: | Core/General |
| Versions: | Frequency of Occurrence | ||
| Related Issues: | |||
| Environment: | Attachments: | ||
| Description: | Ifa user has new voicemail, a SIP Notify with MWI is sent to the user phone. However, if asterisk is restarted or reloaded, the Sip Notify with MWI is re-sent. If there are a large number of users with unread voicemail, then there is a flood of Notify with MWI messages at restart or reload. At restart or reload, the do_monitor function in chan_sip.c should update all peers with the current state of new/old messages before deciding to send a new Notify with MWI. | ||
| Comments: | By: Matt O'Gorman (mogorman) 2006-04-26 11:35:44 umm this really doesnt seem like a major bug, or even a bug at all, more like a feature request. how many phones do you have that this notify traffic is a "problem"? Mog By: bill (bnwcd) 2006-04-26 12:25:15 We have test scenario that simulates around 1000 phones. By: Terry Wilson (twilson) 2006-04-26 17:34:35 I've actually had it take down a proxy server before. Of course, I had around 2000 registered peers on a single box (don't ask) at the time... I believe you can have the same problem if you have qualify turned on as well. By: Serge Vecher (serge-v) 2006-05-04 16:32:40 long time ago, there was a fix on the bug tracker, that spaced out registrations in chan_sip after a reload. I can't find the bug # now, but perhaps the code for this feature could be derived from current reload code. By: Olle Johansson (oej) 2006-05-08 08:21:57 Well, this is not registrations... I think we need to send it out, since we have no clue on what the client knows about the state. But spacing out makes sense. By: Olle Johansson (oej) 2006-06-01 12:26:28 Since this is not a bug, but an effect of the current design, I will mark this as post 1.4 so we can take time to solve it properly By: Olle Johansson (oej) 2006-06-27 13:00:21 Added a space of 100 ms between those initial pokes at reload. Committed to svn trunk #36178 By: Olle Johansson (oej) 2006-06-27 13:01:11 But this was about the mwi's - which is a different thing... Oh, well. Let's look into that then... Sorry. By: Andrey S Pankov (casper) 2006-06-27 19:27:37 oej: why not to apply initial pokes fix to 1.2? Sometimes I can make Asterisk 1.2.9.1+ deadlocked due to those pokes on reload... By: jmls (jmls) 2006-10-31 12:52:09.000-0600 can this be closed, or will it also be applied to the 1.2 series ? By: Olle Johansson (oej) 2006-11-13 14:52:45.000-0600 For 1.2 we will have to be a bit careful and just space out the notifys. For 1.4, we have to disable MWI checks for peers that are not online and possibly add some more checking for NAT status, registration status and such. By: Olle Johansson (oej) 2007-01-23 02:53:19.000-0600 Fixed in 1.4 | ||