summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* plug leakbug25297Matthias Radestock2012-11-161-1/+1
|
* fixmeMatthias Radestock2012-11-161-1/+2
|
* refactor stop_laterMatthias Radestock2012-11-161-18/+15
| | | | | | - s/stop_later/stop - since it's always called with Reason=normal there is no point passing that around / storing it
* Merge bug25295Simon MacMullen2012-11-161-4/+5
|\
| * always monitor queues from which we are (possibly) delivering messagesbug25295Matthias Radestock2012-11-151-4/+5
| | | | | | | | | | so that stats associated with these queues are cleared out from the channel's process dictionary when these queues disappear.
* | refactor: extract some helpersMatthias Radestock2012-11-161-14/+16
| | | | | | | | | | ...to make it abundantly clear which queue operations amount to an ack/requeue.
* | refactor: simplify dlx logic top halfMatthias Radestock2012-11-161-30/+20
| |
* | cosmeticMatthias Radestock2012-11-161-2/+2
| |
* | cosmeticMatthias Radestock2012-11-151-24/+13
| |
* | inline (and avoid misleading var name in the process)Matthias Radestock2012-11-151-2/+2
|/
* refactor stats recording in channelMatthias Radestock2012-11-151-35/+29
| | | | | | - specify stats type explicitly rather than inferring it from the structure of the key (yuck!) - s/maybe_incr_stats/incr_stats
* inlineMatthias Radestock2012-11-151-6/+4
|
* Work on older Erlang versionsSimon MacMullen2012-11-142-5/+8
|
* Merge bug25251 (very late!)Simon MacMullen2012-11-131-11/+2
|\
| * Merge in defaultbug25251Simon MacMullen2012-11-1329-602/+885
| |\ | |/ |/|
* | merge bug24915 into defaultMatthias Radestock2012-11-132-6/+14
|\ \
| * \ merge headsMatthias Radestock2012-11-131-0/+3
| |\ \
| | * | test triggering of memory alarmsMatthias Radestock2012-11-131-0/+3
| | | |
| * | | Merge bug25178Simon MacMullen2012-11-130-0/+0
| |\ \ \ | | |/ /
| * | | And, err, this thing isn't needed any more.bug25178Simon MacMullen2012-11-131-49/+0
| | | |
| * | | Explain whySimon MacMullen2012-11-131-0/+4
| | | |
| * | | rabbit_node_monitor traps exits, so shutdown allows us time to flush anyTim Watson2012-11-131-0/+1
| | | | | | | | | | | | | | | | pending file system operations before the process dies
| * | | roll back to (almost) default - keeping the extra error information whenTim Watson2012-11-132-32/+9
| | | | | | | | | | | | | | | | corrupt cluster status files are found
| * | | merge default into bug25178Tim Watson2012-11-134-24/+47
| |\ \ \ | | |/ /
| * | | Fix idiocySimon MacMullen2012-11-091-1/+1
| | | |
| * | | Take two monitors into the shower.Simon MacMullen2012-11-081-8/+23
| | | |
| * | | ExplainSimon MacMullen2012-11-081-0/+8
| | | |
| * | | Make the app marker into a gen_serverSimon MacMullen2012-11-082-15/+44
| | | |
| * | | Further cleanSimon MacMullen2012-11-081-1/+1
| | | |
| * | | reduce distance to defaultSimon MacMullen2012-11-081-18/+16
| | | |
| * | | introduce a noop process at the head of rabbit_sup's children,Tim Watson2012-11-082-29/+51
| | | | | | | | | | | | | | | | which we now monitor instead of the rabbit application's pid
* | | | tweakbug24915Matthias Radestock2012-11-131-25/+15
| | | | | | | | | | | | | | | | | | | | | | | | - garbage_collect(self()) -> garbage_collect() - rename some funs / msg tags - cosmetics
* | | | Don't hibernate, GC self instead.Simon MacMullen2012-11-121-6/+3
| | | |
* | | | Start background_gc earlierSimon MacMullen2012-11-121-1/+1
| | | |
* | | | Remove define, change message, silence dialyser.Simon MacMullen2012-11-121-4/+3
| | | |
* | | | SpecsSimon MacMullen2012-11-121-0/+9
| | | |
* | | | cast not call.Simon MacMullen2012-11-122-10/+9
| | | |
* | | | Use background_gc when the memory alarm goes off.Simon MacMullen2012-11-122-2/+9
| | | |
* | | | Decrease geometrically as well.Simon MacMullen2012-11-121-1/+2
| | | |
* | | | Profiling suggests that garbage_collect/1 is in fact far cheaper than ↵Simon MacMullen2012-11-121-12/+5
| | | | | | | | | | | | | | | | processes/0, so just GC all waiting processes. And therefore reduce the frequency we call this with.
* | | | Only GC idle (waiting) processes. Filter tiny processes from the list (to ↵Simon MacMullen2012-11-121-8/+10
| | | | | | | | | | | | | | | | make generating the list more efficient when there are huge numbers of processes). Probabilistic process selection.
* | | | Background GC.Simon MacMullen2012-11-123-0/+105
| |/ / |/| |
* | | Merge bug25260Simon MacMullen2012-11-084-17/+40
|\ \ \
| * \ \ merge default into bug25260bug25260Matthias Radestock2012-11-081-6/+6
| |\ \ \ | |/ / / |/| | |
* | | | cosmeticMatthias Radestock2012-11-071-7/+7
| | | |
* | | | merge bug25260 into defaultMatthias Radestock2012-11-071-2/+2
|\ \ \ \ | |_|/ / |/| | |
| | * | fix testsMatthias Radestock2012-11-081-1/+1
| | | |
| | * | pass q records instead of names between recovery phasesMatthias Radestock2012-11-082-9/+10
| | | | | | | | | | | | | | | | | | | | That way we can guarantee that the qpids we are talking to in the start phase are the same as in the recovery phase
| | * | defer final phase of queue recoveryMatthias Radestock2012-11-083-15/+37
| |/ / | | | | | | | | | | | | | | | until routing recovery has completed That way dead-lettering during the final phase has all the correct routing in place
| * | Apply TTL after DLX, during recovery we want to have a working DLX when ↵Simon MacMullen2012-11-071-2/+2
|/ / | | | | | | messages are expired.
View Lorry Controller Status