It's really misleading to print "-- plugins running" when there are no plugins running.

rebase with master

merge ee8aecef6eae into junk

fix variable handling in rabbitmq-server preventing background node startup from working (rebase foo)

remove message about re-installing windows service after plugin changes

be mindful of windows line endings whilst parsing pid files

on windows, die in the correct manner when node name is already in use

revert to expected RABBITMQ_NODE_ONLY behaviour on unix

cosmetic

rationalise rabbit_misc:terminate/1/2 with rabbit_misc:quit/1

fix variable handling in rabbitmq-server preventing background node startup from working

manually merge 3c20079f0601

rework e4649f81c3a6 commit history for renaming

change the cluster nodes functions to rely on the cluster status file only

+ some other fixes based on the fact that now we firmly assume that the
cluster status is as up to date as it can be.

merge c1060fa2e555 into junk

fix variable handling in rabbitmq-server preventing background node startup from working (rebase foo)

remove message about re-installing windows service after plugin changes

be mindful of windows line endings whilst parsing pid files

on windows, die in the correct manner when node name is already in use

revert to expected RABBITMQ_NODE_ONLY behaviour on unix

cosmetic

rationalise rabbit_misc:terminate/1/2 with rabbit_misc:quit/1

fix variable handling in rabbitmq-server preventing background node startup from working

move rabbit_control to the same naming scheme as rabbit_plugins

merge 7dae0f4505ab manually

rework commit history for bug24792

change rabbitmqctl `join_cluster', add `recluster' and `change_node_type'

also, remove `cluster'.

Description for the commands:
* recluster <clusternode>
Instruct an offline node to try to cluster with the nodes provided when
it comes back online. With "cluster" I mean cluster directly, only if the
other nodes think the node is in the cluster already.
This command is useful if we bring a node down and then change the cluster
while it is offline.
* change_node_type (ram|disc)
Changed the node to the provided node type. The change will fail if the
user is trying to create a standalone ram node.
* join_cluster clusternode
Same as before, but only takes one node as "discovery" node. We think
this is better since we don't really make any attempt to verify that
all the provided nodes are in the same cluster anyways.

simplify `rabbit_misc:get_options/4'

Incidentally, it is also much more efficient (not that it matters).

I will probably rename it to something like `parse_arguments'.

`CommandOpts' and `GlobalOpts' could be one, but I think that `GlobalOpts'
is really useful since adding the global options to each single command
is a pain. We could, obviously, put them together in a tuple, but then
it's not that different from two separate arguments.

It's really misleading to print "-- plugins running" when there are no plugins running.

We can't now write the serial record when we declare the exchange; an exchange can become federated at any time.

moved functions around in rabbit_mnesia

Move parameter validation into its own module.

Unused variable warning

Prevent DOWN messages from overtaking initial catchup

merge default

How long have we supported pluggable exchange types now?

Document policy info item.

Remove decorator:route/2.

TODO--: Update behaviour

Track a member's own last_pub in members_state

fix variable handling in rabbitmq-server preventing background node startup from working

First sketch of a policy mechanism.

oops

add test for `rabbit_misc:get_options/4'

change to how `get_options/4' handles flags

Now it doesn't complain on flags that should not be there, it simply eats
only the flags relevant to the command.

don't wait (nearly) forever for mirrored_supervisor migrations
...and reduce wait time for migrations we expect to fail, so we keep
the overall run time short.

merge bug24362

added checks to `rabbit_misc:get_options/3'.

See comments on top of the new definition (now its arity is 4).
I still need to update the tests.

start supervisor children as, er, supervisors, not workers

expand rabbit_sup API with starting of children that are supervisors
and use that API throughout the code where appropriate

cosmetic

cosmetic

remove message about re-installing windows service after plugin changes

be mindful of windows line endings whilst parsing pid files

on windows, die in the correct manner when node name is already in use

add missing spec

get a node to clean up after old incarnations of itself
...in case other nodes don't do so in time or, if they themselves
stop/restart, ever.

revert to expected RABBITMQ_NODE_ONLY behaviour on unix

cosmetic

rationalise rabbit_misc:terminate/1/2 with rabbit_misc:quit/1

move rabbit_control to the same naming scheme as rabbit_plugins

split rabbit_plugins into client and server side modules

rename bootstrap_environment to prepare_plugins; fix while_process_is_alive activity call

merge default

store more info about the cluster on disc, check other nodes before clustering

Now the `cluster_nodes.config; doesn't store a "config" anymore, but all
the information we need about the cluster nodes. This file is updated
whenever a new node comes up. Moreover, this file should be *always*
present, and it's set up in `rabbit:prepare/0'.

Now that we have this file, the various functions regarding the status
of the cluster (`all_clustered_nodes/0', `running_clustered_nodes/0;, etc.)
can "work" even when mnesia is down (and when the rode is a ram node).
Given the assumption that the status file is up to date, with the status
of the cluster (before the node went down if the node is down) a lot of
operations become easier.

I'd like a review at this point, the best thing is probably to diff directly
with default since a lot of previous commits are not relevant anymore.

The most noticeable changes compared to default:

* `join_cluster' works differently. The nodes provided will be used to
"discover" the cluster, and are not used to determine whether the node
should be ram or not. The node will be a disc node by default, and
can be initialized as RAM with the `--ram' flag. The old `cluster'
command is preserved, adapted to work with the new functions while
preserving similar semantics (we decide on whether the node is disc or
not by looking at the list of nodes provided).
* `force_cluster' has been removed.
* The `join_cluster' operation will fail if:
- The node is currently the only disc node of its cluster
- We can't connect to any of the nodes provided
- The node is currently already clustered with the cluster of the nodes
provided
* On default restart RAM nodes try to connect to the nodes that were first
given by the user, and are not aware of the changes that might have
occurred in the cluster when they were online. Since the cluster status
is kept updated on disk now, the RAM node will be aware of changes in the
cluster when restarted.
* Before starting up mnesia, the node contacts the nodes it thinks it's
clustered with, and if the nodes are not clustered with the node anymore
the startup procedure fail. We fail only when we know for sure that
something is wrong - e.g. it won't fail it it doesn't find any online
node

Things to do:

* Implement `uncluster'/`leave_cluster' to kick out a node from a cluster
from another node - this is easy.
* Implement something like `change_node_type', since given how `join_cluster'
works it is not possible right now.
* Rewrite the tests regarding to clustering.
* Think hard about what can go wrong regarding the cluster status file and
the relevant functions. The stuff in `rabbit_upgrade' is particularly
worrying, and I need to make sure that things will work when upgrading
rabbitmq, by reading old file or upgrading them.
* Split `init_db/4' in various functions. We have much stronger assumptions
now, for example we should never need to reset or wipe the schema in
there. In general it's an ugly function, expecially the optional upgrade
part
* Probably something else...

remove ref to old define; document terminate/1 vs quit/1

remove vestigial ERROR_CODE define and dangling TODO from rabbit_misc

preserve original 'wait on any app' functionality

rabbit_control waits as long as process is alive

Merge default

Flatten API for getting the SSL peer name

fix leave_cluster, update rabbit_app.in

change function to find out about cluster nodes

See comment above `check_mnesia_running/1' in `rabbit_mnesia'. In short,
the function to find out about all/running nodes in the cluster do not
work if mnesia is down and the node is a ram node, while we assumed
that they would always work. Moreover, the functions to get the cluster
disc nodes require mnesia to be up.

Make the list of plugins look a bit more like the boot steps.

"Activating RabbitMQ plugins ..." is no longer particularly true or useful to report here. And clean up some trailing spaces.

merge e1311dd42eba into bug24362

merge f6d1609212f9 into bug24362

merge default into this branch

skip non-tests that were causing failures

reseting when clustering, `reset' now disconnects correctly

Also, I check that we're not joining a cluster we're already in.

added `join_cluster', deprecated `cluster, changed the structure of the node config

Merge default

remove the bug 24919 stuff

merge bug24919

run all test suites regardless of failures

removed `rabbit_mnesia:force_cluster'

I'll also have to remove the `Force' parameter from `init_db', fixing
bug 24815.

remote TODO and extraneous branches from rabbit-server

remove the subscription to the schema table

Since dump_tables won't work with the schema table (and works different to what
I had understood anyways) I don't think it's worth it to subscribe to the
table, and it's probably better to just modify the config file through
rabbit_node_monitor.

refactor

dumping schema table on changes in ram nodes

merge default

`proplists:get_bool/2' does not do what I think it did, using `lookup/2'

check for a new pid when registered children die in mirrored_supervisor_tests

rabbitmqctl uses rabbit:await_startup/0

migrate application handling code to app_utils

do not kill the child explicitly

for bug24362 - assert that worker name is unregistered by the time we get a 'DOWN' notification

Include the SSL options in the ssl record, so it can be examined by e.g., rabbit_auth_mechanism_ssl

Merge default

Oops

rename start_cold to boot

revert deleted comment explaining why we're sleeping for 100ms on startup

do not overwrite RABBITMQ_PLUGINS_DIR on windows

apply startup changes to windows service batch utility

fix windows batch file code path handling during cold starts

forgot to change the arity of init_db in the export

re-added a boolean parameter to init_db to prevent upgrading

remove the fun argument in `rabbit_mnesia:init_db/3'.

merge bug24914

attempt to make rabbitmq-server.bat behave like it's unix cousin

ensure that all dependent applications are stopped when shutting down the broker without halting the emulator

move application lifecycle (loading) into one place - rabbit_misc

plugin environment bootstrap and introspection code moved to rabbit_plugins

rabbitmq-server uses start_cold

start and start_cold both defer to rabbit_misc to calculate application load order

rabbit_prelaunch only checks for duplicate node names

add rabbit:start_cold/0 without relying on release/boot-script

pmon is very much hipe-worthy.

Merged bug24889 into default

don't try to disconnet from nodes on `rabbit_mnesia:reset/1' and Force=true

This is the behaviour we had before, and trying to disconnect all the necessary
nodes without connecting to mnesia first is tricky (the previous solution was
very brittle, e.g. the tests break because we disconnect the coverage node).

cosmetic

cosmetic

Merge bug24729 (again)

Fiddle with comments

fix the rabbitmqctl node detection

Merge bug 24901

disconnect from nodes when Force=true.

Right now I'm using a quite ugly hack to be sure to exclude the rabbitmqctl
node, better ideas are welcome.

cosmetic

Don't use an untrappable kill.

Switch to monitoring the overall supervisor.

remove io:format

`rabbit_mnesia:reset/1' did not get the running cluster nodes correctly

Junk bug 24362

Revert 9d59000c4a44, 38eb4652c7e3, f6d1609212f9, e1311dd42eba (timeout changes for mirrored_sup tests)

Revert aa8a9e4d1bb2 (extra debugging for mirrored_supervisor)

cosmetic

oops, had left io:format in the code

record only disc nodes in the config

change config to a {[node()], boolean()} tuple.

Merge default

fix `rabbit_mnesia:create_cluster_nodes_config/2'

Add join_cluster, chang the way cluster config works

The arguments of `join_cluster' are a set of nodes to use to discover the disc
nodes present in the cluster. The old `cluster' command now emits a warning
and is actually `join_cluster'. I still need to make the configurations update
when a new disc node joins the cluster.

The node's cluster config now stores the discovered nodes and whether
the current node is on disc or not. This is less hackish then the previous
solutions (the node is on disc if it's included in the list of cluster nodes
in the config), and makes some code in `rabbit_upgrade' nicer. Note that this
might change in the future anyways, since it doesn't make much sense for a disc
node to have that kind of config.

Things that I'm not sure of:
* What does `force' mean in this context. Right now I left the
`force_cluster' command in place, and it calls
`rabbit_mnesia:force_cluster' as before. The only difference between force
and not force is that when `force = true' and the node cannot discover the
disc nodes, it simply uses the provided nodes. I'm not sure this is
desirable, but since `force' means "make it work even if the nodes are
offline" (at least in this context), I think it's the only way out. The
differences between force/non force in `rabbit_mnesia:init_db/4' remain the
same as before.
* What do we do with old configurations. Right now I simply check how
they're formed and convert them to the new format, but I'm not sure if it's
better to update the file or not. Again this might change since in the
future I'll have to set up a way to update the config file anyways.

Fix for R12B-3 that was broken by bug 24196.

fix typo and opt checking

use pp flag isntead of re-writing the flag

A bit more debugging information on why we might get an error back from the delegate.

Removed unused types, added dummy -spec to silence compiler

merge default

Check that options are not used in the wrong place

merge default

Merge bug 24196.

Log errors when children fail to start.

Fix typo in debian control file

Merge heads

Added tag rabbitmq_v2_8_2 for changeset 4136141f70b9

another attempt to track down the mirrored_supervisor_tests CI failure
waiting forever is all very well...except when we *expect* a failure

cosmetic

let's wait forever
...so we can investigate what is going since this fails in CI even
with a 1 minute timeout

bug: code_change/3 should return {ok, State}

Merge bug24911

another attempt at pacifying the CI gods

Display things that are strings as strings rather than terms, since Erlang's string-detection algorithm breaks on non-ASCII characters. Assume that ~p's use of quotation marks was a deliberate aesthetic choice.

properly ignore dynamic children that return 'ignore' from init/1
thus preventing an explosion on supervisor shutdown when we would
otherwise trip over unlinking an 'undefined' pid.

Merge bug24909

Report error better in case disk free check fails

Correct disk free checking command on windows

Share the exchange scratch space

bump the mirror_supervisor_tests wait time to reduce ci failure rate

Merged bug24866 into default

Merge bug 24790

One less substitution. Pity RHEL and SUSE also differ gratuitously.

Move the Debian init script to its natural place

Move the remaining common init script to RPMs

Merge bug24884

Allow stopping stopped; and starting started broker

Consistency and description

Add an abstraction for reading the scratch

Rework event serialisation thing so that decorators get a say too. It's a bit fiddly since we have to ensure that just because a decorator wants serials, doesn't mean everything else has to get them...

lsb init script for debian

Add comment explaining why ack() and state() exist.

Update the msg_id_ack in set_length.

Merge bug24196

Merge bug24848

Cosmetic. Ish.

Merged bug24885 into default

Merge bug24877

755 is wrong, this should not be executable. Somehow it ends up not executable anyway, but we shouldn't say it is...

I don't think we need a licence header, and it's intimidating. Rephrase a bit too.

AckMsgs -> AckRequired in master:purge

merge bug24885

Damn sleep in Makefile...

Make rabbit_variable_queue:dropwhile tail recursive

merge emile qc changes

Clearer var naming in BQ:dropwhile. Fix in amqqueue_process:drop_expired_messages.

Also, corrected the dropwhile-related comments in rabbit_backing_queue.

For what concerns the fix in drop_expired_messages, I was wrongly generating
acktags for dead messages not to be dead-lettered.

QuickCheck follows dropwhile API updates

Merged bug24833 into default

Increase buffer size for the port driver to match that used by the underlying socket.

Merge heads

Junk bug24866

Merge bug24860

Reorder exception contents