Monitoring QoS stats

[Anon Anon]

It seems like this should be straight-forward but is there a way to monitor the QoS recieved stats and then alert if they drop below a certain threashold? 

 

In the Hub there is the statitics button that looks like every ten minutes it takes a snapshot of QoS sent/recieved per minute. I would like to capture this as a QoS measurement and alert if it drops below a certain threashold. 

[Anon Anon]

I think you may be referring to the button in the data_engine, and I would probably not rely on that value. I think it is reporting the QoS throughput over a relatively short interval, which means it will vary a lot. It may be difficult to come up with a meaningful threshold based on that number. It would be more useful to average the QoS throughput over something like five minutes. It will still vary but not nearly as much. Of course, this is not quite as simple.

 

You can get statistics from the hub that tell you how many messages are in a queue and how many have already been sent out of the queue since it was connected. If you have the values from the previous check, you can calculate the average QoS messages sent and received per minute or second.

 

The longer the polling interval, theoretically the less variation in values under normal load. But five minutes is sort of a sweet spot because a lot of probes typically collect QoS every five minutes by default. Within a five-minute window there will be a lot of variation because each probe is on its own timing, but each five-minute window will include almost all QoS objects currently being collected.

 

This is something I have done before and need to do again...hopefully soon...

[Anon Anon]

Thanks Keith - I am referring to the statistics button on Hub General tab - The collection interval seems to be every ten minutes. I was just hoping for a way to capture this as QoS

[Anon Anon]

Ah, you are talking about the graphs rather than just the point-in-time measurement. Now I see. As you probably realize, those graphs report total messages sent and received, not just QoS. But usually QoS dwarfs all other types of messages by so much that the rest have little impact on the statistics.

 

It looks like the get_perf_data callback has those statistics. The callback is actually documented in the hub whitepaper. According to that, the only argument that is used is the day argument. You just specify they day of the month to get the stats for that day. The data is overwritten after a month, so specifying the day makes it unique.

[RobPol]

I have attached lua script that my help. It the get_stats callback on the data engine to get msg/minute and alerts if it drops below a certain floor. It also sends QoS on each cycle. I fire it from the nas scheduler every few minutes. I got this from the forum and tweaked it a bit for my own use.

 

Regards,

Robin.

 

Edited because my original had actual cut the send QoS off the end

[Anon Anon]

Thank you very much Robin