Hi,
I haven't had time to look at this again until now...
I restarted qpidd the on the (current) backup server, and after that, I
can't reproduce the issue. I guess this means it was indeed related to
replication. Unfortunately, I never managed to enable protocol trace so
I could verify it. How is it done, again? (Please excuse my
forgetfulness.) I tried setting QPID_LOG_ENABLE to "trace:Protocol"
before starting the process, but it produced no output. Maybe I might
have done something on the broker instead, but turning on trace for
everything didn't seem like a good idea...
Some questions that remain to be answered:
1. If this was a backup broker issue, why did it affect only one queue?
2. Is there any way to configure the broker(s) or queues so that there
won't be a complete lock-up like this? I'd rather loose the
replication if there is an issue, but I don't really want to disable
it altogether, if you know what I'm saying...
- Toralf
On 26/05/16 18:16, Gordon Sim wrote:
On 25/05/16 13:53, Toralf Lund wrote:
On 24/05/16 16:41, Gordon Sim wrote:
On 24/05/16 15:28, Toralf Lund wrote:
On 24/05/16 13:22, Gordon Sim wrote:
On 24/05/16 12:20, Gordon Sim wrote:
if you run:
qpid-config queues
and:
qpid-send -q
I think that should allow us to see if there is any queue near its
limit.
qpid-config queues says
Queue Name Attributes
========================================
pgs.bocs.status --replicate=all --lvq-key=qpid.subject --argument
qpid.ha-uuid=1935a318-a308-49ff-bf56-3c618599349a --argument
qpid.last_value_queue=True --argument qpid.browse-only=True
You only have one queue?
I have only one that's supposed to get the messages I'm talking about
here, but perhaps you wanted data for *all* queues? A full list is
included below.
The version of qpid-send I have doesn't seem to have a "-q" option. I
did try
qpid-send -a pgs.bocs.status -m 1
Doh! I'm very sorry, I meant to write `qpid-stat -q`. That gives us
the depths of the queues in question.
Of course. I've used qpid-stat -q before, but I somehow didn't make the
connection... You will find the output at the end of the message.
So it doesn't look like any of the queues are being flow controlled. I
see you have replication enabled, is it actually active? Tne other
possibility is that the completions are being withheld by the broker
because it hasn't replicated. Are there any errors or warnings in any
of the broker logs?
Can you get a protocol trace for the sender in question? That would
confirm that the issue is the broker not sending the completion in the
first place. Then we can try and figure out why that might be.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
