12

Every now and then my drive fills up with temp files from SOLR, and I can't figure out why that is happening. Here's what I know:

I have a 50gb partition and my analytics index has numDocs:43,287,203 and maxDoc:43,316,981. We currently have 15 cores that live on that partition. The analytics index is usually around 30gb but every so often it fills the drive. We restart the SOLR services to dump the temp disk space usage, and that seems to fix the issue. This is a basic install of SOLR (we haven't done any tweaking to the configs). We are on SOLR 5.4.0.0.

Questions:

  1. Is there a way to limit the merges it makes so it won't eat up my disk space?
  2. If not, what are the recommended disk space sizes for these merges? I can get more disk space if I need it, but I need to know how much to ask for.

Thanks for your help!

  • What merges are you doing? The question is not clear on that – Richard Seal Oct 25 '16 at 16:34
  • 1
    I'm definitely not an expert on this. But according to this answer, there is no way to limit them. stackoverflow.com/questions/13167767/… – Mark Cassidy Oct 25 '16 at 16:36
  • We're not doing any merges. The merges that I'm talking about are whatever SOLR does behind the scenes. I'm not sure what the merges are doing, but the best that we can tell, it's during some of those merges that the large growth is happening. – Keith VanderVeen Oct 25 '16 at 16:40
  • What Solr version are you on? – Klaus Petersen Oct 25 '16 at 17:17
  • I updated the question with the current version. – Keith VanderVeen Oct 25 '16 at 17:27
8

This can be a number of things, but it sounds like the merge processes might be the thing. Look into this https://cwiki.apache.org/confluence/display/solr/IndexConfig+in+SolrConfig for more information.

To limit disk space, you should use very LOW merge factors (segmentsPerTier in tieredmergepolicy - mergefactor in LogByteSizeMergePolicy). It will make indexing slower though (might be tremendously slower).

If you have a growing index, you need more disk space. You need to use some sort of capacity management to foresee growth and align this with document growth and expectations of this to know what to ask for, but with indexes and growing contentsize taken into account, you just need to ask for a lot more than you expected.

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.