[Spce-user] mr8.5.1 grafana view problem (kam_concurrent_calls) after kamailio inbound timeout

[Matthias Hohl]

Hello,

 

i found out a funny bug today:

 

We had yesterday night a short high load peak, and after this high load peak
where kamailio got an 1 minute inbound timeout my Grafana diagram shows me
more concurrent calls as exist on the server.

The view is generated with this:

 

SELECT "kam_concurrent_calls" FROM "autogen"."kamailio" WHERE $timeFilter
GROUP BY "host"

Also 

SELECT "kam_usrloc_regusers" FROM "kamailio" WHERE $timeFilter

 

Is around 30% higher registered users like are registered.

It looks like somewhere in the DB (redis?) are some ghost entries, which
fake the “kam_concurrent_calls” and the “kam_usrloc_regusers” value I see
via Grafana.

 

The “kam_dialog_active” is showing the true value. It is just concurrent
calls and usrloc_regusers which are much higher then in reality.

 

If I do:

# redis-cli -n 4 keys \*

 

There are just the active dialogs shown in the right value, but I can’t find
where I found the kam values and how I can reset them?

 

# redis-cli -n 3 FLUSHDB

 

Could this be a solution for this?

 

 

Mit freundlichen Grüßen,

Matthias Hohl

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sipwise.com/pipermail/spce-user_lists.sipwise.com/attachments/20201001/11f881ee/attachment.html>