lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael McCandless <luc...@mikemccandless.com>
Subject Re: Lucene 6.1.0 index upgrade
Date Fri, 10 Nov 2017 11:41:19 GMT
OK I'm sorry but I think you'll have to re-index; e.g. how norms are
computed and stored in the index has changed, which means even if you could
somehow copy or merge the index over to the new version, the norms would be
incorrect.

Mike McCandless

http://blog.mikemccandless.com

On Thu, Nov 9, 2017 at 6:46 PM, Krishnamurthy, Kannan <
Kannan.Krishnamurthy@cengage.com> wrote:

> Hi Mike,
>
> It doesn’t work either, it throws “Cannot merge a segment that has been
> created with major version 6 into this index which has been created by
> major version 7" message. Let me try it after an upgrade as well. Same code
> below,
>
>         @Test
>         public void mergeUsingCodecReader() throws IOException {
>                   DirectoryReader ir = DirectoryReader.open(lucene610Dir);
>                   SegmentReader air = (SegmentReader)ir.leaves().get
> (0).reader();
>
>                 IndexWriterConfig config = newIndexWriterConfig();
>                 IndexWriter w = new IndexWriter(newEmptyDirectory("target"),
> config);
>                 exception.expect(IllegalArgumentException.class);
>                 exception.expectMessage("Cannot merge a segment that has
> been created with major version 6 into this index which has been created by
> major version 7");
>
>                 w.addIndexes(air);
>                 w.forceMerge(1);
>                 w.close();
>         }
> Thanks,
> Kannan
>
> On 11/9/17, 5:07 PM, "Michael McCandless" <lucene@mikemccandless.com>
> wrote:
>
>     Maybe try the addIndexes method that takes CodecReader[] instead?  It
> will
>     be somewhat slower, since it fully reads and writes the index, but it
>     should work across major versions?
>
>     Mike McCandless
>
>     http://blog.mikemccandless.com
>
>     On Thu, Nov 9, 2017 at 3:59 PM, Krishnamurthy, Kannan <
>     Kannan.Krishnamurthy@cengage.com> wrote:
>
>     > Greetings.
>     >
>     >
>     > We are currently using Lucene 6.1.0, our typical index size is ~
> 200GB and
>     > we are planning to move to Lucene 7.0.0.
>     >
>     >
>     > Our daily indexing workflow currently uses IndexWriter.addIndexes()
> to add
>     > an index(this will be on Lucene 7.0.0) to a copy of the of the 200GB
> index.
>     > Further it calls IndexWriter.forceMerge(1) to create a single
> segment. The
>     > addIndexes() now doesn't support different versions of index.
>     >
>     > We are facing an issue in upgrading to Lucene 7.0.0 due to this
>     > limitation. We couldn't afford to do full reindex as there about 50
>     > different indexes ~ 200GB each.
>     >
>     >
>     > We tried using the LuceneUpgrade tool to upgrade our index to latest
>     > version, however the IndexCreatedVersionMajor is still
> LUCENE_6_1_0.major,
>     > This is preventing us calling IndexWriter.addIndexes() post upgrade.
>     >
>     >
>     > I am wondering what are our options to upgrade without doing a full
>     > reindex and to continue to use IndexWriter.addIndexes() post upgrade.
>     >
>     > The code snippet below shows the different api calls we make.
>     >
>     >
>     >
>     > @Test
>     > public void addIndexWith7SegmentToDirectoryOnUpgradedSegment()
> throws
>     > IOException {
>     > IndexUpgrader upgrader = new IndexUpgrader(lucene610Dir); // lucene
> 6.1.0
>     > directory 200 GB
>     > upgrader.upgrade();
>     >
>     > IndexWriter w = new IndexWriter(lucene610Dir,
> newIndexWriterConfig());
>     > exception.expect(IllegalArgumentException.class);
>     > exception.expectMessage(
>     > CoreMatchers.containsString("Cannot use addIndexes(Directory) with
>     > indexes that have been created by a different Lucene version"));
>     > w.addIndexes(lucene700Dir); // lucene 7.0.0 daily index
>     > w.forceMerge(1);
>     > w.close();
>     > }
>     >
>     > Let me know if you need more details.
>     >
>     >
>     > Many thanks
>     >
>     > Kannan.
>     >
>     >
>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message