> I'm not sure how to minimize the impact of performing a re-index (or new > index creation), that will take 48 hours (using oak 1.4). I mean, I don't > want to block other indexes update.
Is this a SegmentNodeStore based setup or DocumentNodeStore based? The reindexing log would have some stats around time spent in indexing and time spent in text extraction. Can you check whats the part which takes most time. If its text extraction then you can reduce the time spent in that via using Pre-Extraction support [1]. This allow extracting text before hand and then using that at time of actual indexing Changing the "indexing lane" should help but is tricky to get right and something we are improving currently OAK-6246 and OAK-5553 > indexes won't be updated. On the other hand, it seems that using the > *reindex-async* flag (see OAK-1456 > <https://issues.apache.org/jira/browse/OAK-1456>) could do the trick. I This mode is useful for property index as in the end it removes the async flag and makes the index synchronous which would cause issues for lucene based index Chetan Mehrotra [1] https://jackrabbit.apache.org/oak/docs/query/lucene.html#text-extraction On Tue, Jun 6, 2017 at 9:02 PM, Alvaro Cabrerizo <topor...@gmail.com> wrote: > Hello, > > I'm not sure how to minimize the impact of performing a re-index (or new > index creation), that will take 48 hours (using oak 1.4). I mean, I don't > want to block other indexes update. > > First, we have set the value of async as *fulltext-async* for the new > index. I guess, that at least, all the indexes managed by the *async* lane > <http://jackrabbit.apache.org/oak/docs/query/indexing.html#indexing-lane> > will not be affected (please, confirm if I'm right). Then we try to > minimize the impact on the fulltext-async lane. According to OAK-5553 > <https://issues.apache.org/jira/browse/OAK-5553> there isn't much we can do > while the indexing process is active for the new index, as the rest of > indexes won't be updated. On the other hand, it seems that using the > *reindex-async* flag (see OAK-1456 > <https://issues.apache.org/jira/browse/OAK-1456>) could do the trick. I > mean, setting reindex-async=true to the new index will allow other indexes > (in the same lane) being updated while it is being populated? If that is > true, we could create the index with that flag and then remove it. > > Regards.
