Average response time on Haproxy Dashboard

[Tek_Chand]

Hello Team,

I am using ELK 6.4.0 and beates (filebeats, metricbeat) 6.4.0. I have exported the dashboard for HAPROXY.

But when i am seeing the Average Response Time for HAPROXY on Kibana dashboard its showing very high i.e approx 121000 ms means around 120 seconds or 2 minutes. Which is not considerable. But we are not facing any issue with our application and sites are loading properly. We also check the HAPROXY logs at back end and did not found any suspicious.

We have checked the qtime, ctime, rtime, ttime in haproxy logs for one day logs but ttime never showing more than 6500 ms. The values of qtime, rtime, ctime are also very low at backend logs.

Can you please let me know how metricbeat calculate the Average response time for HAPROXY? So we can check at our end.

Any assistance will be appreciated.

Thanks.

[jsoriano (Jaime Soriano)]

Hi @Tek_Chand,

Indeed these times look too high, but we take these values directly from haproxy stats backend and according to its documentation they should be milliseconds. Maybe there is something wrong in the dashboards. Could you check if the events stored in Elasticsearch have the same values as the ones seen in the dashboards?

[Tek_Chand]

Hello Jaime,

Thank you for your response.

@jsoriano

according to its documentation they should be milliseconds.

Yes, the average response time is in ms on dashboard. 120 second and 2 minutes are converted values from ms only to understanding the issue.

According to me, In Haproxy stats average response time is not available. May be i am wrong.
In case if i am right then please help me to know that how Metricbeat Haproxy module calculate the average response time?

I will check the events in Elasticsearch and update you.

Thanks.

[Tek_Chand]

Hello Jaime,

@jsoriano

Could you check if the events stored in Elasticsearch have the same values as the ones seen in the dashboards?

Today i have checked the haproxy.stat.response.time.avg values in events at kibana dashbaord for around last one day. The maximum values was 1400 ms. But average response time was very high at haproxy dashboard for same time period.

Please refer the below screenshot:

Please help me to troubleshoot the issue.

Thanks.

[jsoriano (Jaime Soriano)]

@Tek_Chand I have found that this visualization was incorrectly converting from seconds to milliseconds, when the value is already in milliseconds, displaying then values 1000 times higher than expected.

I have opened a PR to fix that https://github.com/elastic/beats/pull/8404

In the meantime you can manually edit this visualization and change data formatter from "seconds to milliseconds" to "milliseconds to milliseconds", like this:

Thanks fro reporting this!

[Tek_Chand]

@Jaime, Thank you for your response as well as for your your efforts.

I have made the changes as you suggested and now average response time is showing good.

Really appreciated thanks once again.