DX Application Performance Management

Hello

We have an APM 10.5 cluster with 10 collectors.

On some colectors we have reached this clamp:

introscope.enterprisemanager.metrics.live.limit = 500000

and increased this value to 700000.

The collectors are virtual servers with:

20 GB RAM

12 GB Heap configured for JVM (-Xms12288m -Xmx12288m)

My question if whther it is allowed to increase this parameter and the rule to calculate the new limit.

thanks and Regards

Antonello

Dear Antonello:

     This is a very good question that should be documented. I asked internally.

    Community:

     Please respond with your insights if you can

   Thanks

Hal German

Hi Antonello,

we recommend and officially support 500,000 metrics per collectors. (Big) IF sufficient hardware and resources are allocated and all the KPIs are in the green, you can exceed that limit (as you already do with 700,000). If you run into issues with GC and heap you might want to increase heap but 12 GB for collectors should be sufficient with improvements made in 10.5.

The most important KPIs are:

• SmartStor duration < 3500ms, ideally <1000ms
• Harvest duration < 3500ms, ideally <1000ms
• GC duration <1000ms, are you using CMS/ParNew or G1GC?
• no "aggregated timeslices" messages in the logs, no collector disconnects or other issues
• all response/processing times under ApplicationTriageMap ok, especially Ontology Engine << 15000ms, Status Engine|Alert Mapping and Status Engine|UVB Mapping

Ciao,

Guenter