Connection: forcibly close stream if close() is called while the connection is blocked

[SimonWoolf]

trafficstars

Because of how rabbit's connection.blocked mode works, if the connection is blocked, the client will not receive a close-ok unless/until the connection becomes unblocked, which may be some time away, or never. Current behaviour just leaves the connection open indefinitely. This instead closes it immediately.

[cressie176]

Hi @SimonWoolf, Thanks for the PR. I see other clients have suffer from the same challenge.

@michaelklishin, it was a while back but do you remember whether and how this issue was address with the Java client? https://groups.google.com/g/rabbitmq-users/c/7cQXglAnWS8

Is there a defacto standard approach that amqplib should follow?

[cressie176]

I've been thinking about this some more and wondering whether supplying an optional boolean "force" to the connection close method might be better. @SimonWoolf if you have time to update the PR then I'll merge. Otherwise I'll take a look, but probably not for a little while.

[SimonWoolf]

I've been thinking about this some more and wondering whether supplying an optional boolean "force" to the connection close method might be better.

to be clear, are you asking for that force arg to apply to the 'destroy() after end()' behaviour change, or to the 'close immediately if blocked' change? (or both?)

[cressie176]

I was thinking of adding the option to connection.close, then passing this through. The function also takes an optional callback which would make this a bit clumsy. Also, something I hadn't considered before, is that this wouldn't help for any non-user invoked closes.

[cressie176]

Some interesting reading here... https://rabbitmq-users.narkive.com/DSpWxIxq/rabbitmq-discuss-blocking-queue-close-connection-by-client

[cressie176]

php-amqplib took a timeout approach - https://github.com/php-amqplib/php-amqplib/pull/609

[kibertoad]

@SimonWoolf Ping? Are you still open to update the PR?