Back to bug 1012852

Who When What Removed Added
Pavel Moravec 2013-09-27 10:39:44 UTC CC pmoravec
Leonid Zhaldybin 2013-09-27 10:50:58 UTC CC lzhaldyb
Justin Ross 2013-09-30 13:37:05 UTC Assignee rhm-maint-list gsim
Justin Ross 2013-09-30 13:37:30 UTC Priority unspecified low
Gordon Sim 2013-10-01 11:29:30 UTC Status NEW ASSIGNED
Doc Text Cause:

In reporting broker statistics or enforcing queue depth limits, only the content was included in the size of AMQP 0-10 messages.

Consequence:

Statistics did not reflect the size of headers in the message. E.g. a message with no body but lots of headers would add nothing to the aggregate size in bytes of a queue onto which it was enqueued. Since for AMQP 1.0 messages the headers and other meta-data included with the message were included, this had the potential to be even more confusing.

Fix:

The size of AMQP 0-10 messages was taken to be the sum of the content and headers segments.

Result:

The statistics whose units are bytes and the queue limits enforced in terms of bytes, now more accurately reflect the logical size of the message and there is less difference between the 1.0 and 0-10 encodings in terms of their overall size.
Gordon Sim 2013-10-01 11:30:36 UTC Doc Text Cause:

In reporting broker statistics or enforcing queue depth limits, only the content was included in the size of AMQP 0-10 messages.

Consequence:

Statistics did not reflect the size of headers in the message. E.g. a message with no body but lots of headers would add nothing to the aggregate size in bytes of a queue onto which it was enqueued. Since for AMQP 1.0 messages the headers and other meta-data included with the message were included, this had the potential to be even more confusing.

Fix:

The size of AMQP 0-10 messages was taken to be the sum of the content and headers segments.

Result:

The statistics whose units are bytes and the queue limits enforced in terms of bytes, now more accurately reflect the logical size of the message and there is less difference between the 1.0 and 0-10 encodings in terms of their overall size. 		Cause:

In reporting broker statistics or enforcing queue depth limits, only the content was included in the size of AMQP 0-10 messages.

Consequence:

Statistics did not reflect the size of headers in the message. E.g. a message with no body but lots of headers would add nothing to the aggregate size in bytes of a queue onto which it was enqueued. Since for AMQP 1.0 messages the headers and other meta-data included with the message were included, this had the potential to be even more confusing.

Fix:

The size of AMQP 0-10 messages was taken to be the sum of the content and headers segments.

Result:

The statistics whose units are bytes and the queue limits enforced in terms of bytes, now more accurately reflect the logical size of the message and there is less difference between the 1.0 and 0-10 encodings in terms of their overall size.

However this also means a slight change to behaviour and queue limits will be reached earlier even with the same messages.
Gordon Sim 2013-10-03 13:38:38 UTC Status ASSIGNED POST
Irina Boverman 2013-10-03 19:08:58 UTC CC gsim, iboverma
Flags needinfo?(gsim)
Gordon Sim 2013-10-04 09:19:40 UTC Flags needinfo?(gsim)
Link ID Apache JIRA QPID-5199
Irina Boverman 2013-10-11 18:41:38 UTC Status POST MODIFIED
Fixed In Version qpid-cpp-0.22-19
Zdenek Kraus 2013-10-22 08:05:38 UTC Flags needinfo?(iboverma)
Zdenek Kraus 2014-01-13 07:59:50 UTC Flags needinfo?(iboverma)
Zdenek Kraus 2014-01-14 13:46:24 UTC Status MODIFIED VERIFIED
QA Contact mrgqe-bugs zkraus
Irina Boverman 2014-02-06 20:58:33 UTC Flags needinfo?(iboverma) needinfo?(iboverma)
Jared MORGAN 2014-09-16 02:20:18 UTC Doc Text Cause:

In reporting broker statistics or enforcing queue depth limits, only the content was included in the size of AMQP 0-10 messages.

Consequence:

Statistics did not reflect the size of headers in the message. E.g. a message with no body but lots of headers would add nothing to the aggregate size in bytes of a queue onto which it was enqueued. Since for AMQP 1.0 messages the headers and other meta-data included with the message were included, this had the potential to be even more confusing.

Fix:

The size of AMQP 0-10 messages was taken to be the sum of the content and headers segments.

Result:

The statistics whose units are bytes and the queue limits enforced in terms of bytes, now more accurately reflect the logical size of the message and there is less difference between the 1.0 and 0-10 encodings in terms of their overall size.

However this also means a slight change to behaviour and queue limits will be reached earlier even with the same messages. 		It was discovered that qpid-cpp was only reporting broker statistics or enforcing queue depth limits based on the size of the AMPQ 0-10 message body, instead of additionally including the header size. This caused statistics to incorrectly reflect the size of headers in the message. For example, a message with no body but many headers would not contribute to the aggregate size in bytes of a queue onto which it was enqueued. Because AMQP 1.0 messages included headers and other meta-data with the message, this issue had the potential to introduce unnecessary confusion. The size of AMQP 0-10 messages is now fixed to take the sum of the content and headers segments into account when calculating the logical size of the message. Statistics and queue limits reported or controlled in bytes now more accurately reflect the logical size of the message, and there is less difference between the 1.0 and 0-10 encodings in terms of their overall size. It is expected that queue limits will be reached earlier due to the correct logical size now being reported and enforced.
John Skeoch 2014-11-09 22:41:43 UTC CC lzhaldyb esammons
Eric Sammons 2015-01-21 12:55:37 UTC Status VERIFIED CLOSED
Resolution --- CURRENTRELEASE
Last Closed 2015-01-21 07:55:37 UTC

Back to bug 1012852