Sitecore 9.2
Solr 7.5
JVM-Memory 1024mb

I have a CM-CD environment for production and I have noticed any time when uploading a CSV via list manager, JVM-Memory spikes to - and maintains a range of - 77%-90%, from 45%-60%. I have implemented a custom contact facet. I have noticed that the standalone servers have no issue with 512mb, but had to increase JVM-memory to 1024mb on CM. The list being uploaded has 1 row and still causes a spike. I am not seeing any error messages on Solr logs. I feel like increasing JVM-memory is a band-aid solution here as our standalone servers are running efficiently with 512mb.

How may I debug this issue efficiently? Would using the List Manager API be useful with custom contact facets?

  • Are you saying that you're running Solr on the same server as your CM instance? May 12 '20 at 17:24
  • Yes @MatthewFitzGerald-Chamberlain May 12 '20 at 19:19
  • Eek. Rolling your own Solr with default settings on the same server as your CM instance is going to be a bad time, especially if you're heavily leveraging xDB/lists/EXM. You should at least get it off the CM box with proper memory/GC settings, but look at just getting it hosted ie by SearchStax May 12 '20 at 19:23
  • We have had the same process for 4+ years with JVM-memory = 512mb. Should adding list and exm functionality make this much of a change? There has been millions of records and still no spikes or issues with Solr/within JVM-memory. Do you still recommend scaling out Solr? May 12 '20 at 19:27
  • To be clear, I would never recommend putting Solr on the CM instance with the default settings in any installation. But putting that aside, yes, the List Manager/EXM is going to put a large load on your indexing and will need to be addressed. May 12 '20 at 19:30

When you upload 1 contact, it tells all the segments and MA plans to take note of the event. This causes a query against all contacts to see if anything has changed and if any contact needs to move.

I production, our JVM is at 3.6GB with 10GB xDB index size.

  • Is this specific to CM instances? In standalone versions, I don't see this behavior. May 12 '20 at 17:16
  • Should see it in both. Is there a number of contacts difference between the two environments? In prod we have 1 mil contacts.
    – Chris Auer
    May 12 '20 at 17:18
  • My production has just over 1million+ contact records in dbo.xdb_shard0 and shard1.Contacts. I have resotred shard0 and shard 1 into a standalone and tried uploading a csv via list manager and minimal spikes (45-55%) were shown. This standalone has 512mb in JVM-memory. May 12 '20 at 19:15
  • Do you have the same segmented lists and MA plans in your standalone? You would need to bring Core/Master/Web/MA and Reference Data databases to copy those down also.
    – Chris Auer
    May 12 '20 at 19:26
  • 1
    Will bring them down now. Just a note, we are not using MA at all. May 12 '20 at 19:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.