Hi all,
I think we've got to the bottom of this now, the good new is that it
looks like it's not a qpid issue per se.
We've got a pretty big federated network and we've recently looked into
partitioning it up better to make management and administration easier
unfortunately..... during that exercise, and despite having quite a
detailed transition plan, it looks like a number of the original queue
routes were left in place instead of being cleaned up when the new paths
were established.
If I didn't laugh I'd cry :-(
I'd be interested in hearing of how others manage large deployments, the
tools are good for what they are and QMF gives a lot of good stats, but
visualising large networks is still pretty hard. I really must crack on
and get my QMF2 web UI finished, it's my spare time project but it'd
probably have saved us a load of grief had I actually had it in a state
I was happy to release :-(
On 29/05/12 23:17, Fraser Adams wrote:
Hi all,
we appear to be seeing some fairly large numbers of message
duplicates, we've had a few other issues that have caused us to be
more observant of what's passing through the system so we're not sure
if this is recent of we've simply not noticed before.
We've got quite a heavily federated system with producers delivering
to multiple brokers (generally three) providing increasing levels of
consolidation.
when we see dupes we've seen two three or four instances of an item
but no more and when they occur they appear to happen pretty much
immediately after each other a few tens of milliseconds apart.
The routes are "push" queue routes (using qpid-route -s) and
delivering to headers exchanges. There's nothing fancy about the
route, so they are using normal unreliable links as per the default on
federated routes.
can anyone shed any light on this sort of behaviour? Any thoughts on
what might be causing it and how to minimise the issue.
BTW we've not proven that it is definitely any issue with qpid or the
routes so we're investigating message counts and sampling messages at
each hop, but we want to cover off all angles so any thoughts on
scenarios that may cause significant dupes would be really useful.
MTIA,
Frase
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
