Hi!

I'm using version 2.2.0 (Erlang 14B on windows) with a shovel between two rabbit servers. When the shovel plugin looses it connection to the remote rabbit server (tested by stopping the rabbit service at the other end) the number of used Erlang processes seems to grow steadily each time the shovel tries to reconnect.

These processes are not reclaimed when I start the remote rabbit server back up. Noticed this when it had been running for a while and rabbit stopped when reached max number of Erlang processes available.

.. Just downloaded v 2.4.0 and without doing too much testing, this version seems to behave properly. When the shovel tries to reconnect, the Erlang process count rises by 30, but drops back by 30 before a new reconnect attempt is performed.

Am I right in thinking there was a bug in 2.2.0 that has been fixed, or do I need to do further testing? :) Just did a 2-minute experiment...

regards,
?yvind

Search Discussions

  • Matthew Sackman at Apr 1, 2011 at 3:54 pm
    Hi ?yvind,
    On Fri, Apr 01, 2011 at 05:51:04PM +0200, ?yvind Tjervaag wrote:
    I'm using version 2.2.0 (Erlang 14B on windows) with a shovel between two rabbit servers. When the shovel plugin looses it connection to the remote rabbit server (tested by stopping the rabbit service at the other end) the number of used Erlang processes seems to grow steadily each time the shovel tries to reconnect.

    These processes are not reclaimed when I start the remote rabbit server back up. Noticed this when it had been running for a while and rabbit stopped when reached max number of Erlang processes available.

    .. Just downloaded v 2.4.0 and without doing too much testing, this version seems to behave properly. When the shovel tries to reconnect, the Erlang process count rises by 30, but drops back by 30 before a new reconnect attempt is performed.

    Am I right in thinking there was a bug in 2.2.0 that has been fixed, or do I need to do further testing? :) Just did a 2-minute experiment...
    You are correct - the bug was reported to us a while ago and has since
    been fixed. I think it might have been fixed in 2.3.0/2.3.1 too, but
    it's certainly fixed in 2.4.0.

    Matthew
  • Øyvind Tjervaag at Apr 1, 2011 at 4:48 pm
    Ah, better go and upgrade some production servers then!

    Thanks :)

    /?yvind
    On Apr 1, 2011, at 5:54 PM, Matthew Sackman wrote:

    You are correct - the bug was reported to us a while ago and has since
    been fixed. I think it might have been fixed in 2.3.0/2.3.1 too, but
    it's certainly fixed in 2.4.0.

    Matthew
  • Matthew Sackman at Apr 1, 2011 at 4:55 pm

    On Fri, Apr 01, 2011 at 06:48:37PM +0200, ?yvind Tjervaag wrote:
    Ah, better go and upgrade some production servers then!
    Be careful. There is, sadly, a bug in 2.4.0 which means that upgrades
    from 2.3.1 or earlier will fail to automatically upgrade the on-disk
    format of msgs will fail in some cases.

    No data will be lost - the upgrade code is very careful to take backups
    before it starts, and will say where in the logs.

    If you have durable queues with msgs on disk that you need to preserve
    then I do not recommend upgrading to 2.4.0. In this case, 2.3.1 is a
    much better idea.

    If you do not have any persisted msgs in durable queues, or you don't
    care about losing any such messages, then upgrading to 2.4.0 is fine,
    though you'll have to manually wipe out your mnesia dir (the logs will
    tell you where this is) after upgrading.

    Best wishes,

    Matthew
  • Øyvind Tjervaag at Apr 1, 2011 at 5:04 pm

    On Apr 1, 2011, at 6:55 PM, Matthew Sackman wrote:
    Be careful. There is, sadly, a bug in 2.4.0 which means that upgrades
    from 2.3.1 or earlier will fail to automatically upgrade the on-disk
    format of msgs will fail in some cases.

    No data will be lost - the upgrade code is very careful to take backups
    before it starts, and will say where in the logs.

    If you have durable queues with msgs on disk that you need to preserve
    then I do not recommend upgrading to 2.4.0. In this case, 2.3.1 is a
    much better idea.

    If you do not have any persisted msgs in durable queues, or you don't
    care about losing any such messages, then upgrading to 2.4.0 is fine,
    though you'll have to manually wipe out your mnesia dir (the logs will
    tell you where this is) after upgrading.

    Best wishes,

    Matthew
    I see. I'll stop all my producers first, let the messages drain off, and then do the upgrade :)

    Thanks for the heads up!

    /?yvind

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
grouprabbitmq-discuss @
categoriesrabbitmq
postedApr 1, '11 at 3:51p
activeApr 1, '11 at 5:04p
posts5
users2
websiterabbitmq.com
irc#rabbitmq

People

Translate

site design / logo © 2022 Grokbase