[ https://issues.apache.org/jira/browse/LUCENE-7976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16427642#comment-16427642 ]
Erick Erickson commented on LUCENE-7976: ---------------------------------------- This is coming together, here's a preliminary patch. It has nocommits and several rough spots/hard-coded numbers, code commented out etc. I'm putting it up in case anyone interested in this wants to take a look at the _approach_ and poke holes in it. Please raise any concerns but also please don't spend a lot of time on the details before I wrap up things I _know_ will need addressing. Current state: 0> refactors a bit of findMerges to gather the stats into a separate class as that method was getting quite hard follow. I haven't made use of that new class in forceMerge or expungeDeletes yet. 1> forceMerge and expungeDeletes respect maxMergedSegmentSizeMB 2> regular merging will do "singleton merges" on overly-large segments when they're more than 20% deleted docs. 20% is completely arbitrary, don't quite know the correct numbers yet. That handles the case of a single-segment optimize not getting merged away for a long time. 3> forceMerge will purge all deleted docs. It tries to assemble max-sized segments. Any segments where the live docs are larger than maxMergedSegmentSizeMB get a singleton merge. 4> fixes the annoying bit where segments reported on the admin UI are improperly proportioned 5> expungeDeletes now tries to assemble max sized segments from all segments with > 10% deleted docs. If a segment has > 10% deleted docs _and_ it's liveDocs > maxMergedSegmentSizeMB it gets a singleton merge. What's left to do: 1> more rigorous testing. So far I've just been looking at the admin UI segments screen and saying "that looks about right". 2> Normal merging rewrites the largest segment too often until it gets to max segment size. I think it also merges dissimilar-sized segments too often. 3> compare the total number of bytes written for one of my test runs between the old and new versions. I'm sure this does more writing, just not sure how much. 4> allow forceMerge to merge down to one segment without having to change solrconfig.xml. 5> perhaps refactor, findMerges, forceMerge and findForcedDeletesMerges to make use of common code. 6> ???? > Make TieredMergePolicy respect maxSegmentSizeMB and allow singleton merges of > very large segments > ------------------------------------------------------------------------------------------------- > > Key: LUCENE-7976 > URL: https://issues.apache.org/jira/browse/LUCENE-7976 > Project: Lucene - Core > Issue Type: Improvement > Reporter: Erick Erickson > Assignee: Erick Erickson > Priority: Major > Attachments: LUCENE-7976.patch, LUCENE-7976.patch > > > We're seeing situations "in the wild" where there are very large indexes (on > disk) handled quite easily in a single Lucene index. This is particularly > true as features like docValues move data into MMapDirectory space. The > current TMP algorithm allows on the order of 50% deleted documents as per a > dev list conversation with Mike McCandless (and his blog here: > https://www.elastic.co/blog/lucenes-handling-of-deleted-documents). > Especially in the current era of very large indexes in aggregate, (think many > TB) solutions like "you need to distribute your collection over more shards" > become very costly. Additionally, the tempting "optimize" button exacerbates > the issue since once you form, say, a 100G segment (by > optimizing/forceMerging) it is not eligible for merging until 97.5G of the > docs in it are deleted (current default 5G max segment size). > The proposal here would be to add a new parameter to TMP, something like > <maxAllowedPctDeletedInBigSegments> (no, that's not serious name, suggestions > welcome) which would default to 100 (or the same behavior we have now). > So if I set this parameter to, say, 20%, and the max segment size stays at > 5G, the following would happen when segments were selected for merging: > > any segment with > 20% deleted documents would be merged or rewritten NO > > MATTER HOW LARGE. There are two cases, > >> the segment has < 5G "live" docs. In that case it would be merged with > >> smaller segments to bring the resulting segment up to 5G. If no smaller > >> segments exist, it would just be rewritten > >> The segment has > 5G "live" docs (the result of a forceMerge or optimize). > >> It would be rewritten into a single segment removing all deleted docs no > >> matter how big it is to start. The 100G example above would be rewritten > >> to an 80G segment for instance. > Of course this would lead to potentially much more I/O which is why the > default would be the same behavior we see now. As it stands now, though, > there's no way to recover from an optimize/forceMerge except to re-index from > scratch. We routinely see 200G-300G Lucene indexes at this point "in the > wild" with 10s of shards replicated 3 or more times. And that doesn't even > include having these over HDFS. > Alternatives welcome! Something like the above seems minimally invasive. A > new merge policy is certainly an alternative. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org