Summary: | ASTERISK-14307: busylevel does not work with realtime | ||
Reporter: | Tim King (timking) | Labels: | |
Date Opened: | 2009-06-12 11:34:01 | Date Closed: | 2011-06-07 14:08:23 |
Priority: | Minor | Regression? | No |
Status: | Closed/Complete | Components: | Channels/chan_sip/DatabaseSupport |
Versions: | Frequency of Occurrence | ||
Related Issues: | |||
Environment: | Attachments: | ||
Description: | The code that checks for busy_level is within a lookup for its name that excludes realtime support "because it would break rtautoclear". Certain aspects of call limit also affected, but call limit (mostly?) works because it's done in update_call_counter. Suspect busy_level should be checked in update_call_counter as well. ****** ADDITIONAL INFORMATION ****** in sip_devicestate if ((p = find_peer(host, NULL, FALSE, FINDALLDEVICES, TRUE))) { ^^^^^ ... else if (p->call_limit && p->busy_level && p->inUse >= p->busy_level) /* We're forcing busy before we've reached the call limit */ res = AST_DEVICE_BUSY; | ||
Comments: | By: Tilghman Lesher (tilghman) 2009-08-24 21:45:19 I have no idea what you want here. Could you upload a candidate patch, please? By: Tim King (timking) 2009-08-25 03:24:19 I can't fix this myself as I don't understand the issue that is documented in sip_devicestate, which is: /* If find_peer asks for a realtime peer, then this breaks rtautoclear. This * is because when a peer tries to autoexpire, the last thing it does is to * queue up an event telling the system that the devicestate has changed * (presumably to unavailable). If we ask for a realtime peer here, this would * load it BACK into memory, thus defeating the point of trying to clear dead * hosts out of memory. This documents the reason why busylevel does not work with realtime peers given the current implementation. The third argument to find_peer is false which means it doesn't look for realtime peers. But it would be nice if busylevel *did* work with realtime peers [or at the very least it was documented elsewhere the restriction]. By: Leif Madsen (lmadsen) 2009-09-22 08:37:29 Honestly, this issue is in such a position that it needs to be discussed on the asterisk-dev list first before moving forward here. I'm going to suspend this issue for now, and the reporter is welcomed to reopen the issue after discussing on asterisk-dev. Thanks! |