introduce stand-in indexer #10
Conversation
Fails to circumvent Sealed WorkQueue exception but highlights that the simple act of trying to index suffices to trigger it.
configurable analyzer used in @ClassBridge(analyzer = ...)
Also reuses fixed set of jobs in the scheduler.
For the cron expression the indexer just notes if it is blank. Can purge not-included types in case they used to be included.
* now respects omero.search.include_actions * still ignores omero.search.max_partition_size but now avoids unnecessary indexing * minor refactor of getWhen into FTI2.Event enum * handle unknown entityType values
LimitTokenCountAnalyzer would be an alternative should we ever wish to set this kind of property to a lower value. In Hibernate Search 3.4.2's Lucene 3.1.0 the code now reads, DEFAULT_MAX_FIELD_LENGTH = MaxFieldLength.UNLIMITED.getLimit();
|
To see more of its activity can add to <logger name="ome.services.fulltext.FullTextIndexer2" level="DEBUG"/> |
|
Currently testing via ome/omero-build#6 |
|
Now running into just a couple of integration test failures, mostly intermittent. Probably nothing major, will investigate. Consistent failure is of Update: Naturally, now it passes after cleaning out the DB. Probably some bad assumptions. Still some intermittent issues with |
|
|
||
| /** | ||
| * An indexer bean replacing the 5.4 full-text thread with adequate functionality. | ||
| * Exists as a stand-in while Hibernate / Spring upgrade issues remain unresolved. |
There was a problem hiding this comment.
@mtbc : could you extend this a bit with:
- how this differs from the previous indexer (trade-offs, etc)
- and how the implementation functions ("first A, then ...")
There was a problem hiding this comment.
Sure, though I may miss some differences and maybe not in time for m4.
|
The issue with |
|
Ha, |
|
Heh, so: interesting issue: my simple approach to fixing |
Already done in previous code.
|
Green with jars from artifactory after conflict resolution. Merging proactively for M4 and the more significant testing can happen there. |
With Hibernate 3.6 the OMERO 5.4 indexer fails with a Sealed WorkQueue exception. This PR introduces a stand-in replacement to index adequately until the problem can be better addressed. See https://trello.com/c/XdwEzrvs/73-indexer-log-hibernate-error.
CLI foreground reindexing (with
--foreground) is not fixed but background should suffice.