Insight SCCM Integration memory usage

[Matthew Dell]

I've noticed a pattern when importing data using the Insight SCCM Integration where memory usage spikes in the host OS and remains at that level until the OS is restarted. A symptom of this is the SCCM import configured in Insight fails to connect to the SCCM DB after a couple of imports and the only thing that solves it is a Jira restart. Other imports continue to function.

Has anyone else experienced similar behavior with this app?

Import details:
-Occurs 1/day
-Affects 2 object types
--One filters the data to import using IQL, bringing the total objects imported/updated from ~16,000 to ~500
--The other imports/updates ~16,000 objects unfiltered.
-Import time is around 1.5 hours using 12 cores

The Jira Xmx is 16GB and the host has 64GB memory. After restarting the server, host memory usage is ~33%. When the SCCM import begins reading data, memory usage on the host moved up to ~38%, when it finishes reading the data, host memory usage is at 84%.

The worrisome part is that the memory seems not to be released. Host memory usage remains at 84% after the import completes and any additional usage is stacked on that until I reboot. I discovered this using monitoring tools and reviewing the insight_import.log; the times of the memory spike and associated actions shown in the log match exactly.

I'm still troubleshooting this, so haven't opened a support case with Riada just yet.  Any thoughts from the community?

 2019-07-16 20:45:00,210 [insight-DefaultImportService-IMPORTS-worker-2-thread] | SCCM Computers>>> Started reading data for configured object type

2019-07-16 22:01:14,193 [insight-DefaultImportService-IMPORTS-worker-2-thread] | SCCM Computers>>> Completed reading data for configured object type

[Matthew Dell · Accepted solution]

I believe this is under control now. We discovered a ballooning condition on the VM, seemingly triggered when the SCCM import ran. We increased RAM allocated to the VM as well as the reservation. In addition we upgraded to v6.3.3 Insight. The memory still spikes but returns to an expected level. So I'd call this a VM configuration issue, rather than an Insight problem. 

[Thomas Deiler]

Dear @Matthew Dell ,

have you changed the GC as recommended by Riada?

So long

Thomas

[Matthew Dell]

Thanks for the quick response, Thomas. Yes, we're using G1GC.

[Thomas Deiler]

Matthew, seems you know what you do - to open that request at Riada is best. Btw, which Insight version do you use?

[Matthew Dell]

I'm using Insight 6.3.2 with SCCM Integration 1.4.2 (latest).
I'm preparing a test with 6.3.3!

Thanks

[Yinon Negev]

Hi Matt.

     Thank you for reporting this.

For the benefit of the community: we are currently investigating this Issue, and hopefully will be able to release an improvement to the SCCM Integration.

Cheers

Yinon

Team Riada