I have some data loading mechanisms in my app that don't adhere to the best practices described here: https://developer.cisco.com/site/axl/learn/how-to/throttling.gsp
Specifically, I'm doing multiple concurrent requests (for c# devs: starting one task another another, then waiting with Task.WhenAll). That usually works even with a bunch of them.. but now I've implemented a generic caching that loads all my cluster data (mostly sql queries).. and that definitely seems to overload AXL even with JSESSIONID reuse. So what happens is that I fire up my tasks, await them, and get a bunch of 503s.
But.. not the 503 examples that are in the guide I linked to above.. but just your average 503 with a html error page saying "service unavailable" in html. So.. why am I not getting any responses like example 2? After all.. that seems to be the most likely cause for the error (there's no way I'm extracting more than 8MB of data or 16MB for all concurrent queries).. the data set I'm extracting is in the KB range.
I'm now adding a throttling layer and I expect that this will solve my issue.. but it doesn't solve the fact that I'm not getting the errors that I should be getting. So why do I get a generic 503 error without any soap errors whatsoever?
No immediate explanation comes to mind...perhaps some resource constraint or throttling mechanism 'above' the AXL implementation code is kicking in, or (perhaps more likely) the AXL throttle response is incorrect or doesn't match the documentation.
I would suggest opening a DevNet Developer Support case so we can take a look at the AXL and system logs, determine where the 503 is coming from, and hopefully update the response to be more helpful/compliant.
It was too late to gather the logs yesterday.. I'll do that shortly so perhaps that'll tell us a bit more about what is going wrong.
Though.. would we even see it in the AXL logs?
Possibly not, I'm anticipating logs from other layers (i.e. Tomcat, DB, auth) might be needed - likely will need engineering guidance for troubleshooting.
I have the same issue, some times CUCM just sent a 503 service unavaible with no explanation. I've checked AXL Logs and found nothing.
I think cisco is a bit little wrong then they say "ExecuteSQLQuery has no limit", I suspect that there is some kind of overloading of the AXL service when dealing with multiple sqlquerys.
If you find something usefull on this, please post it here!