| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Only drop-head strategy.
This necessitated the change of rabbit_fifo prefix messages from a tuple
of integers representing the number of returned vs enqueued messages
that have already been processes and thus don't need to include message
bodes in the snapshot to a tuple of lists of the sizes of each message.
This change will have some performance impact as the snaphots will now
be larger than before but as they still won't contain message bodies at
least the sizing is fixed. Decreased the frequency as snapshots points
are prepared so somewhat make up for this.
[#161247380]
|
| |
|
|
| |
References rabbitmq/rabbitmq-cli#292.
|
| |
|
|
|
|
|
|
| |
It joins the club of is_booted/1 and is_running/{0, 1}.
This allows for a CLI command that checks if the node is still
botting.
References rabbitmq/rabbitmq-cli#292.
|
| |
|
|
|
|
| |
Light refactoring
Shortening some lines for the benefit of split buffer users.
|
| | |
|
| |\
| |
| |
| |
| | |
rabbitmq/rabbitmq-management-649-single-active-consumer
Integrate single active consumer with monitoring
|
| | |\
| | |
| | |
| | |
| | | |
Conflicts:
src/rabbit_fifo.erl
|
| | | |
| | |
| | |
| | |
| | | |
For quorum queues. Otherwise dead waiting consumers can be kept in the state
between restarts.
|
| | | |
| | |
| | |
| | |
| | |
| | | |
[#163089472]
References rabbitmq/rabbitmq-management#649
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Add single consumer tag and channel PID to metrics infos in quorum
queues
* Include waiting consumers in consumer count in quorum queues
* Add single active consumer flag in consumer metrics
* Update metrics for newly-promoted single active consumer
[#163089472]
References rabbitmq/rabbitmq-management#649
|
| | |\ \ |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
WIP for quorum queue.
[#163089472]
References rabbitmq/rabbitmq-management#649
|
| |\ \ \ \
| |_|_|/
|/| | | |
Add an upgrade function to remove all default bindings.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Non-durable queues and bindings are not on disk anyways.
This should be changes when moving to feature flags.
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Versions before 3.8 create default bindings for queues in the database.
Since #1721 they are replaced with placeholders only for list operations.
We need to cleanup the bindings to improve binding performance with
older queues.
[#163224049]
|
| |\ \ \
| | | |
| | | | |
Make mandatory based on route data only
|
| | |\ \ \ |
|
| | | |/ /
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Instead of waiting for a mandatory_received message from the queue the
mandatory result is calculated in the channel based on the routing
result only. This may seem like a weakening of the mandatory semantics but
considering that the mandatory_received message is returned _before_ the
message is enqueued and/or persisted in the queue it doesn't actually
open up any further failure scenarios.
[#163222515]
|
| |\ \ \ \
| |_|/ /
|/| | | |
Improve quorum queue deletion
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The default timeout of net_ticktime + 5000 ms was much too long and made
the delete operation in the management UI timeout and not complete
properly. A simple 5000ms works better and this commit also handles the
case where the delete command was committed but the leader never shut
down due to a server being unavailable. This also falls back to forced
deletion.
[#163062078]
|
| | | | | |
|
| | |/ /
| | |
| | |
| | |
| | |
| | | |
Only delete the amqqueue record after successful Ra cluster deletion.
Force delete Ra servers when not able to achieve consensus.
|
| |\ \ \
| |/ /
|/| | |
Remove use of accept_ack
|
| |/ /
| |
| |
| |
| |
| | |
Ranch no longer calls accept_ack for a transport, which means a couple important operations are missed for new sockets.
Fixes #1835
|
| |\ \
| | |
| | | |
Testing of quorum queues
|
| | |\ \
| |/ /
|/| |
| | |
| | | |
Conflicts:
src/rabbit_fifo.erl
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
rabbitmq/rabbitmq-server-1825-single-active-consumer-in-qq-leaks-consumers
Clean consumers properly in QQ when single active consumer is on
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Necessary when single active consumer is on, otherwise only the single
active consumer shows up in the command line.
|
| | | | |
| | | |
| | | |
| | | | |
For single active consumer.
|
| | | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | | |
For single active consumer. Otherwise they still show up in the
management plugin. Worse, a consumer from a dead channel can end up
being the single active consumer, which stops deliveries.
References #1825
|
| | | | |
|
| | | |\
| |_|/
|/| | |
|
| |\ \ \
| |/ /
|/| | |
Fixed years in copyright
|
| |/ / |
|
| |\ \
| | |
| | | |
Update to sysmon_handler 1.1.0
|
| | | | |
|
| | | |
| | |
| | |
| | | |
Also ensure that sane defaults are used for sysmon_handler if not set by the user
|
| |/ / |
|
| | |
| |
| |
| | |
Discovered while doing acceptance. References #1812.
|
| | |\
| |/
|/| |
|
| | |
| |
| |
| |
| | |
This is a short term solution, the generated config helpers/framework is
due for revisiting.
|
| | | |
|
| |\ \
| | |
| | | |
Add sysmon-handler to RabbitMQ
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fixes #952
Update for rabbitmq-sysmon -> sysmon-handler rename
Use hex.pm package for sysmon_handler, copy schema into rabbit.schema
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
rabbitmq/rabbitmq-server-1799-single-active-consumer-in-qq
Single active consumer
|
| | |\ \ \ |
|
| | | | | | |
|