summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
...
| * | | Exclude consumers suspected as down from the consumer countdcorbacho2020-09-281-1/+4
|/ / /
* | | maintenance_mode_SUITE: Skip leader election checkJean-Sébastien Pédron2020-09-252-7/+36
| | | | | | | | | | | | | | | ... in `quorum_queue_leadership_transfer` if some `rabbit_fifo` machines have different versions.
* | | per_user_connection_channel_limit_SUITE: Wait for channel/connection to ↵Jean-Sébastien Pédron2020-09-251-13/+13
| |/ |/| | | | | | | | | | | | | terminate This is a follow-up to commit f95bb4a4b1468a05c69202a2e0e03d063bc248de which did the same thing for just a few cases. Now all similar cases are handled the same.
* | Update rabbit_definitions:apply_defs specs to feature returned errorsMichael Klishin2020-09-241-5/+5
|/
* Make it easy to close all connections on a nodeMichael Klishin2020-09-241-1/+8
| | | | | from a test. HTTP API test suites would greatly benefit from this.
* Erlang 21.3 is no longer supported, drop the CI badgeMichael Klishin2020-09-241-1/+0
|
* Merge pull request #2441 from rabbitmq/erlang-22Michael Klishin2020-09-242-561/+561
|\ | | | | Test with Erlang 22.3 and 23.0
| * Test with Erlang 22.3 and 23.0erlang-22Philip Kuryloski2020-09-232-561/+561
|/ | | | Replace the GitHub Actions workflow for 21.3 with 22.3
* Make maintenance mode tests that involve QQs conditionalMichael Klishin2020-09-231-8/+41
|
* Fix one more timing-sensitive per_user_connection_channel_limit suite testMichael Klishin2020-09-231-6/+22
|
* Merge pull request #2447 from rabbitmq/qq-peek-commandMichael Klishin2020-09-226-17/+152
|\ | | | | Quorum queue peek
| * rabbit_fifo: type the messages lqueueqq-peek-commandkjnilsson2020-09-221-1/+1
| | | | | | | | To pass dialyzer
| * Test invariant: peek must not drain the queuedcorbacho2020-09-221-0/+2
| |
| * Quorum Queue Peek commandkjnilsson2020-09-226-17/+150
|/ | | | | | Allow peeking of messages into a quorum queue. This uses a dedicated aux command to peek at a message at the given 1-indexed position in the queue.
* Skip another mixed versions testkjnilsson2020-09-211-0/+9
| | | | as it cannot be made reliable.
* per_user_connection_channel_limit_SUITE: Wait for channel/connection to ↵Jean-Sébastien Pédron2020-09-211-5/+5
| | | | | | | | | | | | terminate ... `most_basic_cluster_connection_and_channel_count` testcases. `amqp_channel:close()` is a synchronous call in the AMQP client. However we verify the number of remaining channels/connections on the server side. This explains the fact that there might be a window of time where the server-side resources might not be completely cleaned up when the client-side calls return.
* quorum_queue_SUITE set lower aten poll intervalkjnilsson2020-09-211-1/+3
|
* Switch per_user suite to query feature flagkjnilsson2020-09-213-33/+45
| | | | | Instead of the environment to avoid running incompatible mixed versions tests.
* Merge pull request #2448 from rabbitmq/reclaim-qq-memoryMichael Klishin2020-09-203-4/+48
|\ | | | | Quorum queue: API to reclaim quorum memory
| * Change a log message to be more accuratereclaim-qq-memoryMichael Klishin2020-09-182-3/+3
| | | | | | | | | | A full sweep GC does not always decrease process heap size in the moment.
| * Quorum queue: API to reclaim quorum memorydcorbacho2020-09-162-2/+46
| | | | | | | | | | Operator can choose to force a garbage collection and flush of the WAL on a specific quorum queue
* | Ignore some tests when for mixed versionskjnilsson2020-09-183-10/+24
| | | | | | | | As they are not able to run.
* | Merge pull request #2450 from rabbitmq/aten-poll-intervalMichael Klishin2020-09-181-0/+9
|\ \ | | | | | | Increase aten poll interval
| * | Increase aten poll intervalaten-poll-intervalkjnilsson2020-09-181-0/+9
|/ / | | | | | | | | | | | | | | | | A longer interval will reduce the likely hood that unnecessary leader elections are triggered in quorum queues due to transient network issues. This increases it from the default of 1s to 5s which increases the time it takes to detect genuine network partitions somewhat. Ra still uses monitors so connection failures will not have to wait this additional time.
* | quorum_queue_SUITE test reliabilitykjnilsson2020-09-172-3/+16
| | | | | | | | For mixed versions.
* | Merge pull request #2449 from rabbitmq/lrb-revert-unintentional-assertionLuke Bakken2020-09-171-7/+7
|\ \ | |/ |/| Revert unintentional assertion
| * Revert unintentional assertionlrb-revert-unintentional-assertionLuke Bakken2020-09-161-7/+7
|/ | | | | | PR #1818 introduced an unintended assertion here. Previously the record was updated with the new queue pid. With the change, it became an assertion.
* Merge pull request #2407 from rabbitmq/qq-reject-publishKarl Nilsson2020-09-1519-499/+4672
|\ | | | | reject-publish strategy support for quorum queues
| * Add system time to unit testqq-reject-publishdcorbacho2020-09-141-1/+2
| |
| * Add a couple of more fields to overviewkjnilsson2020-09-071-1/+4
| |
| * rabbit_fifo: lower min snapshot intervalkjnilsson2020-09-072-19/+17
| | | | | | | | | | As the current default is way to large for queues that never empty but only keep a small backlog.
| * Quorum queue 'Queue TTL' support.kjnilsson2020-09-076-152/+318
| | | | | | | | | | | | The quorum queue will track the timestamp of whenever a consumer was removed and periodically check if the queue should be deleted based on it's expiry configuration.
| * Add some unused fields to rabbit_fifo recordskjnilsson2020-09-071-6/+11
| | | | | | | | | | To allow for future extensions without necessitating a machine version upgrade
| * Don't assert on message countkjnilsson2020-09-071-1/+1
| | | | | | | | | | As depending on which is the current leader this may or may not be returned.
| * rabbit_fifo test changekjnilsson2020-09-071-1/+3
| |
| * rabbit_fifo: convert enqueuerskjnilsson2020-09-071-1/+8
| | | | | | | | To v0 version which has an additional field.
| * fix typokjnilsson2020-09-071-1/+1
| |
| * Better handle version checkkjnilsson2020-09-071-1/+1
| | | | | | | | when the rabbit_fifo:version/0 function is missing.
| * Skip more testskjnilsson2020-09-073-9/+51
| |
| * quorum_queue_SUITE skip 2 node tests for mixed versionskjnilsson2020-09-071-32/+38
| |
| * Skip test that cannot work with mixed machine versionskjnilsson2020-09-071-2/+7
| | | | | | | | Growing shrinking does not necessarily work in mixed versions.
| * Switch to 3 node clusters instead of 2kjnilsson2020-09-073-3/+3
| | | | | | | | As this works better with mixed Ra machine versions
| * Skip test in mixed modekjnilsson2020-09-071-40/+44
| | | | | | | | that cannot pass due to Ra machine versioning restrictions.
| * Test fixeskjnilsson2020-09-073-15/+26
| |
| * Make some test suite 3 node instead of 2kjnilsson2020-09-075-37/+43
| | | | | | | | | | | | | | QQs cannot be declared in two node clusters with different machine versions. This setup should be an anomaly so changing some suites to better represent sensible conigurations. This should also allow some mixed-versions tests to pass.
| * test fixeskjnilsson2020-09-073-3/+22
| |
| * Remove debug log statementkjnilsson2020-09-071-6/+2
| |
| * Implement reject_publish for QQskjnilsson2020-09-077-143/+461
| | | | | | | | | | | | | | | | | | | | | | | | | | | | The reject publish overflow strategy for quorum queues is an inexact implementation that relies on the cooperation of publishing channels. When a channel first wants to publish to a quorum queue it first issues a synchonous register_enqueuer command which will return the current queue overflow state as reject_publish if the queue is full. The queue will also notify any active enqueuers when it reaches the limit but will continue to accept any enqueues it receives after that. Once the queue size goes below 80% of the limit(s) the queue will again notify enqueuers that they can resume publishin inte the queue.
| * Leave release cursors as they arekjnilsson2020-09-071-6/+1
| |
| * rabbit_fifo: convert statekjnilsson2020-09-073-26/+69
| |
View Lorry Controller Status