Hi,

I am using vanilla filesystem for directory_provider. The same index is used for indexing as well as querying. Indexing happens every 10 mins depending on new updates. Indexing is by a manual process triggered by a java scheduler

The system has been running on a test environment for more than 2 months and I do not see any index corruption. However would like to know what are the best practices to be followed to ensure that there is no index corruption

Hi,
in theory the index should' t corrupt even if you killed the JVM at any time. Of course some updates might be lost, but the design of Lucene's indexes should be robust enough to prevent "corruption".

Generally speaking, there's nothing different than any disk-stored recommendation: pick a safe filesystem, no exotic filesystem tuning flags unless you know what you're doing, make sure your disks don't fail.

In practice corruption is of course possible because of bugs as no software is perfect, but it's very unlikely as in this case we're talking about code which is being actively tested by lots of people, in lots of different use cases and workloads.

Might be worth keeping an eye on the Lucene JIRA, that's where you'll find if anyone opens an issue: https://issues.apache.org/jira/browse/LUCENE ; the only couple of times I've seen a corruption issue being reported, it as however resolved at record speed. So another strategy is to make sure you keep updated.