| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
It is not immediately obvious that a 0 value for channel_max means infinite or "no limit". This change will result in the following message being logged:
"negotiated channel_max = 0 (no limit) is higher than the maximum allowed value (2047)"
|
| | |
|
| |\
| |
| | |
Handle generic data structures in priority queue info callback
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The priority queue implementation makes assumptions over the
data format returned by the underlying backing queue
`info` callback.
This causes priority queues to crash if used together with other
modules implementing the `rabbit_backing_queue` behaviour and
returning additional data with the `info` callback.
As we cannot predict the format of the data the backing queue
`info` callback could return, we just replace it with the undefined
keyword.
Signed-off-by: Matteo Cafasso <noxdafox@gmail.com>
|
| | |
|
| | |
|
| |\
| |
| | |
Set two environment variables in the same manner as on Unix systems
|
| |/
|
|
|
|
| |
SERVER_START_ARGS and SERVER_ADDITIONAL_ERL_ARGS from the rabbitmq-env-conf.bat file need to be honored here.
[159602360]
|
| |\
| |
| | |
Add support for syslog server hostname
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| | |
schlagert/syslog version 3.4.3 adds support for specifying a host name for `dest_host`, this PR supports it as well
See this gist for my test procedure:
git@gist.github.com:7ab6348839c3dc7b18b6662f024e089c.git
|
| |\ \
| |/
|/| |
Correct resource limit defaults
|
| | |
| |
| |
| | |
118b448631a699736f1380d7990989c7e61f94ff, back out my changes
|
| |/
|
|
|
|
|
|
|
| |
They were bumped to x10 the values in 118b448631a699736f1380d7990989c7e61f94ff
by mistake.
Kudos to Chris Friesen for reporting this.
Closes #1659.
|
| |\
| |
| | |
Prepare semi-durable binding table once instead of each time a virtual host is recovered
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Before #567 all binding recover functions were executed once on
node restart.
Executing table_filter can be expensive with high number of vhosts
and bindings and effectively should be called only once.
Moved to a separate function, which should be called from the global
recovery function (`rabbit_vhost:recover/0`)
|
| |\ \
| |/
|/| |
Cancel synchronisation when memory alarm is triggered
|
| | |
| |
| |
| | |
[#158828073]
|
| | |
| |
| |
| |
| |
| | |
The cancellation used to block if the memory alarm was set on any of the
slaves during sync. This patch solves it.
[#158828073]
|
| |/ |
|
| |
|
|
|
|
| |
... to keep it under the 80-column limit.
References #1641.
|
| |\
| |
| | |
Additional info for `rabbitmqctl rename_cluster_node` man page
|
| | | |
|
| | |
| |
| |
| |
| |
| | |
The man page currently explains that it only changes the local database.
I have added information about the other steps required to successfully
rename a node.
|
| |\ \
| | |
| | | |
Correctly call rabbitmq-env.bat
|
| |/ /
| |
| |
| | |
Fixes #1642
|
| |/
|
|
| |
References rabbitmq/rabbitmq-server#1640.
|
| |\
| |
| | |
Add code_server cache to prevent code:get_object_code abuse
|
| |/
|
|
|
|
| |
In the case of high connection churn, code:get_object_code is called multiple times to find the Module:additional_authn_params method for the direct connection type. In the case of a missing module, the code_server process can be overrun with messages while checking the code path. This change caches the result so that future calls to a missing or bad module are not as time consuming
Fixes VESC-888
|
| |
|
|
|
|
|
|
| |
At debug level; there already are info messages about accepted and successfully
authenticated and authorized connections.
[#158782152]
[#158782156]
|
| |\
| |
| | |
Pass authentication context to Mod:user_login_authorization/2
|
| |/
|
|
|
|
| |
Part of rabbitmq/rabbitmq-server#1633.
[#158805410]
|
| |\
| |
| | |
(doc) fix typo in mqtt vhost section
|
| |/ |
|
| |\
| |
| | |
Compile on Erlang/OTP 21
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
OTP 21 has a new logging subsystem and extra logging added to the kernel app
to accompany the asynchronous automatic node connection feature. This breaks
this eval command that asks the node for some of its effective paths (e.g.
the plugin dir).
Since we know the set of lines we are interested in, we can filter out
everything else.
Pair: @dumbbell.
Part of #1616.
[#157964874]
|
| |/
|
|
|
|
|
|
|
|
|
| |
OTP 21 deprecated erlang:get_stacktrace/0 in favor of a new
try/catch syntax. Unfortunately that's not realistic for projects
that support multiple Erlang versions (like us) until OTP 21 can be
the minimum version requirement. In order to compile we have to ignore
the warning. The broad compiler option seems to be the most common
way to support compilation on multiple OTP versions with warnings_as_errors.
[#157964874]
|
| | |
|
| |
|
|
|
|
|
|
| |
Previous implementation breaks on OTP 21.
[#157964874]
(cherry picked from commit da011eb94ed79463daf912756dd866f2d777c0c3)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It will never be started on OTP 21.
The problem this piece of code originally was added to solve
is not very relevant now that we use Lager. Arguably connection
rate throttling should be enforced elsewhere, be it Ranch or a proxy/load balancer.
In any case, this particular implementation worked around an error_logger
issue that is no longer problematic for us, and eventually won't be problematic
for the entire ecosystem because OTP 21 introduced a completely new logger.
Cherry-picked per discussion with @dumbbell and @lukebakken.
[#157964874]
(cherry picked from commit a54b44fd8d24af90d8df301658fba4775f957bc8)
|
| | |
|
| |\
| |
| | |
fix default heartbeat in example conf file to 60
|
| |/
|
| |
listed as 600, not the true default of 60
|
| |\
| |
| | |
Use more optimal default memory allocators where available (Erlang 20.2.3 or later)
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We have settled on the following configuration for memory
allocators after testing many different combinations:
+MBas ageffcbf +MHas ageffcbf +MBlmbcs 512 +MHlmbcs 512 +MMmcs 30
They reduce the memory usage and help RabbitMQ reclaim memory,
at the cost of a slight decrease in performance due to an
increased number of memory operations.
We need to start Erlang with these values in order to figure
out whether they are supported. The allocator strategies we
recommend were introduced in Erlang/OTP 20.2.3. The values
can be overriden using RABBITMQ_SERVER_ERL_ARGS.
cc @gerhard
|
| | |
| |
| |
| |
| |
| | |
`erlang:floor/1` is not available in 19.3.
[#156729133]
|
| |\ \
| | |
| | | |
Introduce rabbit_nodes:await_running_count/2
|
| | | |
| | |
| | |
| | | |
[#156729133]
|
