For 4.1.x, by @aaron-seo: introduce a command that would force QQs to take a checkpoint and truncate its segments #13548
Conversation
|
@aaron-seo @kjnilsson note that due to how Otherwise the checkpoint is not actually taken in the test, even if the @kjnilsson I have rolled back the |
| {QName, {error, Err}} | ||
| end | ||
| end | ||
| || Q <- rabbit_db_queue:get_all_durable_by_type(?MODULE), |
There was a problem hiding this comment.
this will include queues that do not have any members on the current node but the force_checkpoint_on_queue will only ask any local member to force a checkpoint.
| {ok, Q} when ?amqqueue_is_quorum(Q) -> | ||
| {RaName, _} = amqqueue:get_pid(Q), | ||
| rabbit_log:debug("Sending command to force ~ts to take a checkpoint", [QNameFmt]), | ||
| rpc:call(Node, ra, cast_aux_command, [{RaName, Node}, force_checkpoint], ?FORCE_CHECKPOINT_RPC_TIMEOUT); |
There was a problem hiding this comment.
there is no need to do an rpc here as this function always addresses the quorum queue member on the node it runs on. is this the intent? to force a checkpoint on a single member if there is one locally (there are no checks to ensure that).
| def help_section, do: :replication | ||
|
|
||
| def description, | ||
| do: "Forces checkpoints for all matching quorum queues" |
There was a problem hiding this comment.
no it will only force a checkpoint on the local member of the matching queues, if there is a local member.
|
When it comes to triggering a checkpoint for all replicas on the given node vs. all replicas, period, I guess the right thing to do is to do it on all replicas. I'll look into it. |
|
We have agreed to finish this most likely after |
michaelklishin
changed the title
This is #13175 by @aaron-seo plus some changes from me that to address the QA feedback from @kjnilsson.