[Azure OpenAI] Metrics data mismatch from Azure Monitor and Ingested Documents

[muthu-mps]

• Currently the Azure OpenAI service has a Timegrain of PT1M. While testing the metrics that is generated in the service and the metrics ingested in Elastic. We see some of the metrics are not 100%.
• This may be because of the delay in the service. We need to verify to fetch complete metrics by increasing the latency and verify if there is a delay in the service.

[muthu-mps]

Updates

I had a sync-up with @zmoog on this issue to trace where exactly it was happening. Thanks @zmoog for your time.

• As an initial step we tried to collect metrics using the beats module which worked fine.
• When we installed the integrations there is a gap in the metrics collection.
• Finally we just disabled the TSDB and tried collecting the metrics which works fine and the metrics matches perfectly with the Azure monitor.

cc - @zmoog

[lalit-satapathy]

Finally we just disabled the TSDB and tried collecting the metrics which works fine and the metrics matches perfectly with the Azure monitor.

While we analyze the TSDB specific issue, we can disable TSDB in short-term on metrics data stream. CC: @tommyers-elastic @andresrc

[lalit-satapathy]

We can keeping it open to track future TSDB enablement, while it does not impact current package.