Martin Ritchie wrote:
On 07/09/2007, Robert Greig <[EMAIL PROTECTED]> wrote:
On 07/09/2007, Rafael Schloming <[EMAIL PROTECTED]> wrote:
The intermittent failures described in QPID-580 could quite possibly
indicate message loss, although it is impossible to know for sure due to
the way the test is constructed.
Also, QPID-573 causes message duplication which, depending on your
application, may be as bad as message loss.
OK. My own feeling is that we should release M2 now and plan to
release an M2.1 (as already discussed) and potentially an M2.0.1 (for
any users who want just the bug fixes without the protocol upgrade
"risk").
M2 (or mostly M2) is being used in production today and has undergone
a fair amount of testing so I think it is reasonable to go ahead with
it. I would rather draw a line in the sand with it, and then get into
a habit of doing much more frequent bug fix and minor improvement
releases.
I would tend to agree. Looking at the JIRAs I don't see any message
loss issues. The worse case is message ordering under load my not be
maintained.
The message ordering problem shows up for me when just publishing three
messages in a row, i.e. not really heavy load, just a quick burst. My
machine may be more prone to it than others though because I test on a
dual CPU system.
Based on my understanding of the problem, I would expect there to be
almost no ordering maintained under continuous load.
I'm also concerned about the correctness of the message selectors when
messages are delivered asynchronously, but that may be my own
misunderstanding of the code. See my comments on QPID-572 for more details.
--Rafael
