summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* yet more testsbug25328Matthias Radestock2012-12-031-2/+21
|
* test moreMatthias Radestock2012-12-031-9/+13
|
* add test for vq:fetchwhileMatthias Radestock2012-12-031-11/+20
|
* WhitespaceEmile Joubert2012-12-031-1/+1
|
* merge default into bug25328Matthias Radestock2012-11-306-58/+81
|\
| * merge default into bug25328Matthias Radestock2012-11-306-58/+81
| |\
| | * return to a simpler, better BQ:dropwhile, and introduce 'fetchwhile'Matthias Radestock2012-11-296-64/+87
| | | | | | | | | | | | | | | | | | | | | ...to cover the remaining required functionality, including the ability to process messages along the way, pass around an accumulator, and get hold of the IsDelivered flag (not needed in our use case but included for similarity with 'fetch').
* | | unbreak qc (hopefully)Matthias Radestock2012-11-301-2/+2
|/ /
* | remove unused state varMatthias Radestock2012-11-301-29/+14
|/ | | | | We were diligently maintaining ack_msg_id, but never actually using it for anything. And it's always been like that.
* merge bug25303 into defaultMatthias Radestock2012-11-2814-274/+269
|\
| * re-merge bug25303 into defaultMatthias Radestock2012-11-2714-274/+269
| |\
| | * Merge bug25323Simon MacMullen2012-11-277-57/+40
| | |\
| | | * mark all messages enqueued in the slave as 'delivered'bug25323Matthias Radestock2012-11-267-57/+40
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | which is a much better than the set_delivered logic, which we can now get rid of. In doing so it also becomes clear that having the 'delivered' flag in the #message_properties is less than ideal. It is mutable and we never bothered updating vq s.t. it sets the flag correctly. So lets get rid of it and add a parameter to bq:publish instead
| | * | cosmeticMatthias Radestock2012-11-271-5/+3
| | | |
| | * | remove cruftMatthias Radestock2012-11-261-1/+0
| | | |
| | * | merge stable into defaultTim Watson2012-11-265-38/+31
| | |\ \
| | | * \ merge bug25309 into stableTim Watson2012-11-265-38/+31
| | | |\ \
| | | | * \ merge stable into bug25309bug25309Tim Watson2012-11-263-25/+42
| | | | |\ \ | | | | |/ / | | | |/| |
| | | | * | recover for accidental merge with defaultTim Watson2012-11-261-1/+1
| | | | | |
| | | | * | Fix a race where we decide on the BQ before writing the #amqqueue{} to ↵Simon MacMullen2012-11-221-4/+5
| | | | | | | | | | | | | | | | | | | | | | | | Mnesia, then the policy changes and we are therefore not notified. We now only make the initial decision after we can be assured we will be informed of subsequent changes.
| | | | * | Carry on if an individual notification fails.Simon MacMullen2012-11-211-2/+2
| | | | | |
| | | | * | Useful guard.Simon MacMullen2012-11-211-1/+1
| | | | | |
| | | | * | Don't try to assert result of {add,drop}_mirror.Simon MacMullen2012-11-211-11/+3
| | | | | |
| | | | * | Make start/stop mirroring into a cast not a call.Simon MacMullen2012-11-212-19/+19
| | | | | |
| | * | | | merge bug25303 into defaultMatthias Radestock2012-11-264-9/+11
| | |\ \ \ \
| | * \ \ \ \ Merged bug25311Emile Joubert2012-11-267-61/+140
| | |\ \ \ \ \ | | | | |_|_|/ | | | |/| | |
| | | * | | | merge bug25324 into defaultMatthias Radestock2012-11-241-1/+1
| | | |\ \ \ \
| | | | * | | | 'pid' is a valid INFO_KEYbug25324Matthias Radestock2012-11-241-1/+1
| | | |/ / / / | | | | | | | | | | | | | | | | | | | | | and some code, namely the clustering tests, rely on that
| | | * | | | merge bug25303 into defaultMatthias Radestock2012-11-235-9/+93
| | | |\ \ \ \
| | | * \ \ \ \ Merged bug25315Emile Joubert2012-11-237-151/+197
| | | |\ \ \ \ \
| | | * | | | | | propagate API changebug25315Matthias Radestock2012-11-221-4/+4
| | | | | | | | |
| | | * | | | | | bq api tweak: don't include remaining message count in fetch/drop resultMatthias Radestock2012-11-226-47/+42
| | | | | | | | |
| | * | | | | | | don't evict messages from RAM when inserting them into pending_ackbug25311Matthias Radestock2012-11-231-10/+7
| | | |/ / / / / | | |/| | | | |
| | * | | | | | refactor: vq:ram_ack_index doesn't need to be a gb_treeMatthias Radestock2012-11-231-12/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | a gb_set suffices
| | * | | | | | merge bug23625 into defaultMatthias Radestock2012-11-234-106/+137
| | |\ \ \ \ \ \
| | | * | | | | | Explainbug23625Simon MacMullen2012-11-221-0/+10
| | | | | | | | |
| | | * | | | | | Don't duplicate name.Simon MacMullen2012-11-221-1/+1
| | | | | | | | |
| | | * | | | | | optimiseMatthias Radestock2012-11-161-2/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | this brings perf roughly on par with default
| | | * | | | | | identify queues by name rather than pid in channel statsMatthias Radestock2012-11-162-67/+86
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - in deliver_to_queues, take the result of rabbit_amqqueue:lookup and use it to a) add entries to a QPid -> QName mapping in the state for all master pids, and b) setup monitors for all master pids - for the stats creation in deliver_to_queues, map the DeliveredQPids to the associated QNames, via the mapping in the State. Note that this will ignore slave pids, hence we only get one stat per QName (which is good). Also, in the event that the master died between lookup and delivery (and the delivery was 'mandatory'), we will not record any stats at all. - in monitor_delivering_queue, which is called by basic.{consume,get} we add to the mapping - in ack/2 we use the mapping to obtain QNames from QPids, and use that in stats. Since a queue may have vanished prior to the ack/reject arriving, we need to handle the case of no entry being present in the mapping for the given QPid. - in record_sent we have the QName anyway, so can just record stats against that instead of the QPid. - in the 'DOWN' handler we use the mapping to determine the QName from the pid, and pass that to erase_queue_stats, which can now remove entries based on the QName. We then remove the entry from the mapping.
| | | * | | | | | propagate API changeMatthias Radestock2012-11-161-1/+1
| | | | | | | | |
| | | * | | | | | merge default into bug23625Matthias Radestock2012-11-162-38/+38
| | | |\ \ \ \ \ \
| | | | * \ \ \ \ \ merge default into bug23625Matthias Radestock2012-11-164-89/+69
| | | | |\ \ \ \ \ \
| | | | * \ \ \ \ \ \ merge default into bug23625Matthias Radestock2012-11-152-38/+38
| | | | |\ \ \ \ \ \ \
| | | | | * \ \ \ \ \ \ merge default into bug23625Matthias Radestock2012-11-151-6/+4
| | | | | |\ \ \ \ \ \ \
| | | | | * | | | | | | | reference queue by name in consumer eventsMatthias Radestock2012-11-151-18/+22
| | | | | | | | | | | | |
| | | | | * | | | | | | | emit names instead of pids in queue process' queue eventsMatthias Radestock2012-11-152-20/+16
| | | | | | | | | | | | |
| | | | | * | | | | | | | restart branchMatthias Radestock2012-11-150-0/+0
| | | | | | | | | | | | |
* | | | | | | | | | | | | expand test coverage to at least two iterations of delta_foldbug25303Matthias Radestock2012-11-281-18/+21
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | previously it had no test coverage at all Also, re-use the same vq for all fold tests, for efficiency.
* | | | | | | | | | | | | tweakMatthias Radestock2012-11-281-19/+14
| | | | | | | | | | | | |
* | | | | | | | | | | | | cosmeticMatthias Radestock2012-11-281-2/+3
| | | | | | | | | | | | |
View Lorry Controller Status