[asterisk-dev] Asterisk freeze and... a catch-22.

[Kirill Katsnelson]

We are running 1.8.1.1 on a prod server that handles only a hundred 
calls an hour or so. The server freezes: sometimes works for 2 weeks, 
sometimes locks up twice daily.

I got backtraces and lock logs. While I was chasing them (2 weeks lost 
because of incorrect instructions to the ops etc), Digium stopped 
accepting bug reports against 1.8.1.1.

Ok, I decided to bite the bulled and upgraded to 1.8.3 rc2... only to 
find it is broken badly enough so we can't use it (I logged an issue 
18729, but that's not a point here). Here's the catch-22: I cannot repro 
in a newer version, as this requires weeks of real life load, and cannot 
normally open an issue in mantis for it.

Now, the lockup is lurking there manwhile, and is the worst of all 
things. If asterisk crashes, it is restarted. But the thing just 
silently stops processing calls, and needs an op to intervene and 
restart. I followed many lock up issues logged in the tracker for the 
past couple months, and am pretty sure this is not among the logged ones.

Question to the list is thus what should I do with the f^H precious 
backtraces I got: still open an issue, although I cannot test 1.8.3 yet, 
or let it pass till 1.8.3 is fixed enough so we can upgrade and see if 
it magically fixed itself, which is unlikely?

  -kkm