Hello!

We're using v2.1.0 on Ubuntu 9.10 with the same versions of the stomp and
amqp adapter and stuck there because some issues our code has with a rewrite
of the stomp adapter that occurred some time soon after 2.1.0.

I'm pretty certain we're bumping into the bug reported here:
http://goo.gl/V4ww2

Last message from that thread:

"I have raised a bug for this problem.

As reported here Matthew feels that the *stomp* client is generating the
wrong form for the payload of a message, and the code verifies this
observation.
On Wed, Feb 02, 2011 at 11:48:19PM +0000, Matthew Sackman wrote:
Ok, so the message payload is being sent through as a list instead of as
a binary. My guess without checking further is this is a bug in the
*stomp* client in the way it constructs the messages which are then sent
further into *Rabbit*, but this may not be the case.
The fix is to send it flattened, of course.

To validate this fix a small testcase is being constructed.
_______________________________________________
rabbitmq-discuss mailing list
rabbitmq-disc...<http://groups.google.com/groups/unlock?_done=/group/rabbitmq-discuss/browse_thread/thread/796e1f0c7ba0551c/fa0f29e51b43d70f%3Flnk%3Dgst%26q%3Drabbit%2Bstomp%2Bcrash&msg=e292a5436aa24076>
@lists.rabbitmq.com
https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss<http://www.google.com/url?sa=D&q=https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss>


"


Any chance of a backport of the fix, please or workaround I might be able to
get in place? :)

Thanks!

crr

Christopher.Rhodes at cloudkick.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20110610/788e76c3/attachment.htm>

Search Discussions

  • Toby Corkindale at Jun 14, 2011 at 12:45 am

    On 10/06/11 23:54, Christopher Rhodes wrote:
    Hello!

    We're using v2.1.0 on Ubuntu 9.10 with the same versions of the stomp
    and amqp adapter and stuck there because some issues our code has with a
    rewrite of the stomp adapter that occurred some time soon after 2.1.0. [snip]
    Any chance of a backport of the fix, please or workaround I might be
    able to get in place? :)

    I found it was quite simple to backport the later versions of RabbitMQ
    and plugins - grab the source packages from upcoming ubuntu version
    11.10 and just recompile them into packages for 9.10 - it's mostly
    automatic.
  • Rob Harrop at Jun 14, 2011 at 11:35 am
    What exactly is the problem? The shortened URL seems to be disabled for
    spam ;)

    I'm particularly interested in knowing if there is something in the
    latest version of the adapter that is preventing you from moving. Fixing
    that will be a lot easier than backporting a fix to the pre-2.1.0 era code.

    Rob

    Christopher Rhodes wrote:
    Hello!

    We're using v2.1.0 on Ubuntu 9.10 with the same versions of the stomp
    and amqp adapter and stuck there because some issues our code has with a
    rewrite of the stomp adapter that occurred some time soon after 2.1.0.

    I'm pretty certain we're bumping into the bug reported here:
    http://goo.gl/V4ww2

    Last message from that thread:

    "I have raised a bug for this problem.

    As reported here Matthew feels that the *stomp* client is generating the
    wrong form for the payload of a message, and the code verifies this
    observation.
    On Wed, Feb 02, 2011 at 11:48:19PM +0000, Matthew Sackman wrote:
    Ok, so the message payload is being sent through as a list instead of as
    a binary. My guess without checking further is this is a bug in the
    *stomp* client in the way it constructs the messages which are then sent
    further into *Rabbit*, but this may not be the case.
    The fix is to send it flattened, of course.

    To validate this fix a small testcase is being constructed.
    _______________________________________________
    rabbitmq-discuss mailing list
    rabbitmq-disc...
    <http://groups.google.com/groups/unlock?_done=/group/rabbitmq-discuss/browse_thread/thread/796e1f0c7ba0551c/fa0f29e51b43d70f%3Flnk%3Dgst%26q%3Drabbit%2Bstomp%2Bcrash&msg=e292a5436aa24076>@lists.rabbitmq.com
    <http://lists.rabbitmq.com>
    https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
    <http://www.google.com/url?sa=D&q=https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss>


    "


    Any chance of a backport of the fix, please or workaround I might be
    able to get in place? :)

    Thanks!

    crr

    Christopher.Rhodes at cloudkick.com <mailto:Christopher.Rhodes at cloudkick.com>



    _______________________________________________
    rabbitmq-discuss mailing list
    rabbitmq-discuss at lists.rabbitmq.com
    https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
  • Chris at Jun 15, 2011 at 4:47 pm

    I'm particularly interested in knowing if there is something in the
    latest version of the adapter that is preventing you from moving. Fixing
    that will be a lot easier than backporting a fix to the pre-2.1.0 era code.

    Rob
    Thank you for the response Rob.

    Not sure exactly, but I know that if we move past 2.1.0 and the
    matching stomp plugin things get worse. I'm just the guy that has to
    keep this thing running and not the developer of the stomp client
    pushing data at the rabbit stomp adapter or the clients on the other
    side doing amqp. I'll try again to get the details on what's keeping
    us stuck at this version.

    In the meantime this is what we see when the Queue goes away:

    10:49 < arreyder> =ERROR REPORT==== 15-Jun-2011::15:46:12 ===
    10:49 < arreyder> STOMP error frame sent:
    10:49 < arreyder> Message: "Invalid frame"
    10:49 < arreyder> Detail: "Could not parse frame\n"
    10:49 < arreyder> Server private detail:
    {bad_header_key,"destinaSEND"}
    10:49 < arreyder> "detinaSEND"
    10:51 < arreyder> here's the one from just before that one:
    10:51 < arreyder> =ERROR REPORT==== 15-Jun-2011::15:31:11 ===
    10:51 < arreyder> STOMP error frame sent:
    10:51 < arreyder> Message: "Invalid frame"
    10:51 < arreyder> Detail: "Could not parse frame\n"
    10:51 < arreyder> Server private detail:
    {bad_header_key,"destinationSEND"}

    I've verified on the wire (packet captures) that the message headers
    do not look like how they are logged there. They appear correctly
    sent.

    crr
  • Rob Harrop at Jun 17, 2011 at 8:42 am
    Hi Chris,

    This is what you see against 2.1?

    The bug you linked to is not present in 2.1, it was introduced in 2.2
    but is fixed since 2.4.1 (or maybe 2.4.0).

    The log doesn't seem to be saying anything interesting about queues
    crashing, simply that the parser thinks the incoming frame is corrupted.
    In particular, it looks like the header key has a newline in it, which
    is definitely not allowed.

    Do you have a few sample frames that you could send me.

    The best solution to this is to find out why you can't run against
    2.5.0. The parser was much updated to support STOMP 1.1 recently and to
    handle some clarifications around handling of extraneous newlines and such.

    Rob

    Chris wrote:
    Thank you for the response Rob.

    Not sure exactly, but I know that if we move past 2.1.0 and the
    matching stomp plugin things get worse. I'm just the guy that has to
    keep this thing running and not the developer of the stomp client
    pushing data at the rabbit stomp adapter or the clients on the other
    side doing amqp. I'll try again to get the details on what's keeping
    us stuck at this version.

    In the meantime this is what we see when the Queue goes away:

    10:49< arreyder> =ERROR REPORT==== 15-Jun-2011::15:46:12 ===
    10:49< arreyder> STOMP error frame sent:
    10:49< arreyder> Message: "Invalid frame"
    10:49< arreyder> Detail: "Could not parse frame\n"
    10:49< arreyder> Server private detail:
    {bad_header_key,"destinaSEND"}
    10:49< arreyder> "detinaSEND"
    10:51< arreyder> here's the one from just before that one:
    10:51< arreyder> =ERROR REPORT==== 15-Jun-2011::15:31:11 ===
    10:51< arreyder> STOMP error frame sent:
    10:51< arreyder> Message: "Invalid frame"
    10:51< arreyder> Detail: "Could not parse frame\n"
    10:51< arreyder> Server private detail:
    {bad_header_key,"destinationSEND"}

    I've verified on the wire (packet captures) that the message headers
    do not look like how they are logged there. They appear correctly
    sent.

    crr
    _______________________________________________
    rabbitmq-discuss mailing list
    rabbitmq-discuss at lists.rabbitmq.com
    https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
  • Chris at Jun 15, 2011 at 5:17 pm

    On Jun 14, 6:35?am, Rob Harrop wrote:
    What exactly is the problem? The shortened URL seems to be disabled for
    spam ;)

    Here's the full link to the previous thread that I think is related to
    the bug we are hitting.

    http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/2011-February/011071.html

    crr

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
grouprabbitmq-discuss @
categoriesrabbitmq
postedJun 10, '11 at 1:54p
activeJun 17, '11 at 8:42a
posts6
users4
websiterabbitmq.com
irc#rabbitmq

People

Translate

site design / logo © 2022 Grokbase