On 08/29/2012 10:52 AM, Caesar Holger (KGQI 1) wrote:
Hi Gordon,

4) There is no feature to detect slow/hung consumers
You can get an alert when the backlog of messages for a subscriber
reaches a configured level. That would be indicative of a slow
consumer.
How can I do that? Setting something like
"default-event-threshold-ratio" or "alert_size"? Can we be sure that it
is one single consumer that is too slow on reading and not just the
ensemble of all consumers?

Yes, you can set the point at which an alert is generated on a per queue basis, and each topic subscriber will have their own queue using qpid.alert_size and/or qpid.alert_count when creating the queue (or x-qpid-maximum-message-size, x-qpid-maximum-message-count which are aliases for the same options).

(If you have a queue shared between multiple consumers then the fact that one of them is slow is less relevant; the backing up of the queue can of course still be detected in the same manner).

The alert message will contain the queue name and the depth (message count & aggregate size). That is usually reasonably simple to tie back to a particular topic subscription.

8) SQL92-compliant message selectors as defined in the JMS standard
are not
supported
They are not yet supported by the c++ broker, but the JMS client can
evaluate them on the client side. So logically the feature is there,
though less efficiently implemented than if they were evaluated server
side.
I tried using a message selector in the client using an SQL92 compatible
string from the Oracle JMS API ("NewsType = 'Sports' OR NewsType =
'Opinion'"). However I get the following exception:
"javax.jms.InvalidSelectorException: cannot create consumer because of
selector issue"
Do you have any idea why this is?

No, I don't see why that should be invalid. Its not a particularly helpful error message either... Was there a linked exception tied to that with any more detail.

Anyone on the JMS side able to suggest anything here?




---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to