summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* Fix the remaining memory leakMatthew Sackman2011-05-231-79/+93
|
* Updated documentation, and in the course of writing it, thought up another ↵Matthew Sackman2011-05-232-50/+128
| | | | scenario I wasn't coping with. Fixed. However, not all documented causes of memory leaks are yet fixed in the code
* Remove debug log entriesMatthew Sackman2011-05-232-2/+0
|
* master must broadcast, esp if it doesn't know about the sender. Also rip out ↵Matthew Sackman2011-05-234-67/+40
| | | | the varying priority run_backing_queue* stuff as it turns out it's not needed
* Critical bug fixedMatthew Sackman2011-05-201-1/+1
|
* That's an awful lot of work to solve a potential memory leak...Matthew Sackman2011-05-205-87/+257
|
* ...and the other bits I forgot.Matthew Sackman2011-05-201-2/+3
|
* BQ:idle_timeout => BQ:timeoutMatthew Sackman2011-05-202-10/+10
|
* Merge with defaultRob Harrop2011-05-2015-32/+1714
|\
| * Make log messages prettierMatthew Sackman2011-05-202-8/+11
| |
| * Be more assertive. Slave is the equiv of amqqueue_process, so it's ok here ↵Matthew Sackman2011-05-191-2/+2
| | | | | | | | to know the message is actually a #basic_message and thus contains an id which can be matched against
| * Make slaves explicitly monitor the master (see comments to come in bug)Matthew Sackman2011-05-191-5/+20
| |
| * Merging default to bug23554Matthew Sackman2011-05-191-8/+17
| |\
| * | Improve logging informationMatthew Sackman2011-05-172-2/+4
| | |
| * | Eliminate a race which was found to allow promotion to be non-atomicMatthew Sackman2011-05-171-5/+19
| | |
| * | Add upgrade step to add the mirror pids to the queueMatthew Sackman2011-05-171-0/+15
| | |
| * | Ensure that when a slave gets promoted, it requeues msgs in the same order ↵Matthew Sackman2011-05-171-19/+21
| | | | | | | | | | | | which they were fetched
| * | Add a sleep, to avoid tight spinningMatthew Sackman2011-05-171-0/+1
| | |
| * | Not especially happy with this but can't think of how else to solve this. If ↵Matthew Sackman2011-05-171-2/+7
| | | | | | | | | | | | the queue has mirrors then you may well find that during death you try to contact a dead queue process, because the mnesia table is yet to be updated. In such cases, loop.
| * | Merging default to bug23554Matthew Sackman2011-05-1715-107/+206
| |\ \
| * | | Fixed call to validate_messageRob Harrop2011-05-161-1/+1
| | | |
| * | | Merge with defaultRob Harrop2011-05-1611-127/+338
| |\ \ \
| * \ \ \ Merge with defaultRob Harrop2011-05-034-61/+69
| |\ \ \ \
| * \ \ \ \ Merge with defaultRob Harrop2011-04-1514-30/+1655
| |\ \ \ \ \
| | * | | | | Forgot to undo these bits, and the boot sequence has been changed, so debitrotMatthew Sackman2011-04-133-5/+13
| | | | | | |
| | * | | | | Undo merge from bug24038 which has now become INVALID.Matthew Sackman2011-04-133-133/+59
| | | | | | |
| | * | | | | Work on documentation of haMatthew Sackman2011-04-101-0/+11
| | | | | | |
| | * | | | | Work on documentation of haMatthew Sackman2011-04-102-62/+124
| | | | | | |
| | * | | | | Merging default into bug23554Matthew Sackman2011-04-1019-230/+1932
| | |\ \ \ \ \
| | | * | | | | Additional hilarious witticismMatthew Sackman2011-04-081-1/+1
| | | | | | | |
| | | * | | | | Some initial HA design documentationMatthew Sackman2011-04-083-0/+81
| | | | | | | |
| | | * | | | | Merging bug24038 into bug23554 and hook slave up to the new apis. Overall, ↵Matthew Sackman2011-04-083-73/+137
| | | |\ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | it doesn't save that many lines, but it allows the slave to not duplicate -defines which were also in amqqueue_process, which is well worth the effort
| | | | * | | | | Abstract out mainly timer maintanence functionsMatthew Sackman2011-04-082-35/+118
| | | | | | | | |
| | | * | | | | | Merging bug24039 to bug23554Matthew Sackman2011-04-0814-30/+1503
| | | |\ \ \ \ \ \
| | | | * | | | | | some notesMatthew Sackman2011-04-081-0/+40
| | | | | | | | | |
| | | | * | | | | | Sod it - transactions are too hard to do in mirror queues so don't bother. I ↵Matthew Sackman2011-04-082-235/+44
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | know how to do it, but it's horrifically messy, and the margin is too small
| | | | * | | | | | Well, getting closer. But it's not done yet, and I may have discovered a ↵Matthew Sackman2011-04-082-23/+126
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | rather fatal problem with the whole idea of supporting txns in mirrors anyway in that because of the coalescing going on, there is absolutely no indication of when the BQ finally completes adding the msgs to the queue. Thus the only solution here might be to ban coalescing in this case
| | | | * | | | | | Well txns are still only half in. But I want to go home.Matthew Sackman2011-04-076-44/+135
| | | | | | | | | |
| | | | * | | | | | Grrr. non-maskable-interrupt half way through implementing txnsMatthew Sackman2011-04-072-46/+79
| | | | | | | | | |
| | | | * | | | | | Add BQ:discard, correct BQ:is_duplicate, finally fix the last bits of ↵Matthew Sackman2011-04-076-34/+112
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | immediate delivery, though hopefully in a way which has not leaked through to the lower layers...
| | | | * | | | | | Validate_msg becomes is_duplicate, which makes much more senseMatthew Sackman2011-04-075-26/+33
| | | | | | | | | |
| | | | * | | | | | Merging default to bug23554 (minor debitrot)Matthew Sackman2011-04-0718-192/+1587
| | | | |\ \ \ \ \ \
| | | | | * \ \ \ \ \ merge default to bug23554 (minor debitrot)Matthew Sackman2011-03-2916-183/+637
| | | | | |\ \ \ \ \ \
| | | | | * \ \ \ \ \ \ Merging default into bug23554Matthew Sackman2011-03-2219-234/+1679
| | | | | |\ \ \ \ \ \ \
| | | | | | * \ \ \ \ \ \ Merging default to bug23554 (minor debitrot)Matthew Sackman2011-03-1819-234/+1679
| | | | | | |\ \ \ \ \ \ \
| | | | | | | * | | | | | | whoopsMatthew Sackman2011-03-161-1/+1
| | | | | | | | | | | | | |
| | | | | | | * | | | | | | A different and largely opposite version of 'never'...Matthew Sackman2011-03-161-2/+8
| | | | | | | | | | | | | |
| | | | | | | * | | | | | | Made gm do batching of messages. This has an astonishing performance impact: ↵Matthew Sackman2011-03-162-46/+96
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | if every broadcast msg to the gm results in network activity then performance is low - presumably serialisation of, and network broadcast of small messages is very inefficient. By batching broadcasts and then sending many on a timer, performance is much much higher.
| | | | | | | * | | | | | | Merging default into bug23554Matthew Sackman2011-03-166-92/+86
| | | | | | | |\ \ \ \ \ \ \
| | | | | | | * \ \ \ \ \ \ \ Merging default into bug23554Matthew Sackman2011-03-1417-181/+1573
| | | | | | | |\ \ \ \ \ \ \ \
View Lorry Controller Status