[mythtv-users] MySQL related BE deadlocks - collective wisdom needed
warpme at o2.pl
Fri Jul 29 08:58:45 UTC 2011
This is little off-topic, but maybe it can help You to at least little decrease deadlock's frustration.
5months ago I was on 0.24-fixes and I had very similar situation: MYTH_PROTO deadlocks lading to totally unusable 3 standalone FE + FE as app on desktop. This was so frustrating that I started to be really desperate. I decided give a chance for master.
Rationale to switch was quite simple: even if - by it's active development status - master has many more bugs it can't be worse than 0.24-fixes with it's nasty MYTH_PROTO deadlock.
Another argument was that, assuming my case is case specific, I can't expect dev will be devote enough time to resolve case specific issue - what worse - issue which pops up on code being in maintenance mode.
First impression after switching: features stability is not worse than 0.24, but I still had deadlocks (less frequent but still present).
First breakthrough was when Daniel partially merged rec2 branch. This was night and day for DVB-S2 recording part of my system. In my case this was revolution as now I can reliably multirecord concurrent shows starting at the same time.
Next revolution was when code responsible for MYTH_PROTO was modularized. After this code change, resiliency goes highly up: failure in recording subsystem has much less impact on playback availability. Basically since that time majority of deadlocks which I observe are always related to mysqlclient library and according to current hypothesis are related to DB reconnects.
Taking into account current state of Your's WAF I think You can consider switch to master. Probability of improvement is really high, risks that it will be worse is probably near to zero (there is no worse thing than hung, right ;-p).
This IMHO will be also beneficial for devs as MYTH_PROTO deadlock efforts can be consolidated and directed on single myth version.
I'm using master daily on production system and my current overall impression is that it works better (much better) than 0.24.
Comparing both: I didn't notice any areas where master has noticeable regressions to 0.24 (oh, PIP is one from them) but I see many many improvements.
Speaking by emotions: master is much much more pleasure to use.
Hope it helps
More information about the mythtv-users