What are the recommended values for jira.index.max.reindexes and jira.index.update.bulk.optimization?

[na1s]

Hi everybody.

I have instance with 700 000 tickets. I use these values now

jira.index.max.reindexes=0

jira.index.update.bulk.optimization=0

So I need to reindex every week, to avoid performance problems.

But what values is recommended for my server?

[Nic Brough -Adaptavist-]

Mmm, the best answer is "don't reindex every week" - you should only need to reindex when the index gets broken by a crash or dodgy code or settings, or on certain changes to field setup.  If you feel the need for regular re=indexing, you need to be investigating that need, because it's pointing at something being vastly wrong.

[na1s]

Hi, if I set jira.index.max.reindexes to 0, Jira will never run optimization task, so index will degrade. If I will set it to default value equals 2000 index actions will fail with timeout, so to avoid index degradation I run reindex every week. And this is normal for Jira.:-) So I want to ask, maybe here somebody has highload Jira instance, what params do they use?

[Nic Brough -Adaptavist-]

So don't set the value to 0, it's set to 2000 for a good reason. Stop messing with these values, you're doing absolutely the wrong thing. The default values work for almost all JIRA installations, and in the cases where I do recommend changes because of load or volume, the usual recommendation is to increase them, never reduce them. Your comment "This is normal for Jira" is totally wrong - what is "normal for Jira" is for you to use the defualt indexing values and not break it by reindexing regularly. You need to stop "plastering over the cracks" and look at the actual problem. Why is it timing out?

[na1s]

Optimizing will lock Jira index, so if optimizing will take more than default 30 seconds, jira updates will not be reindexed, because reindex will fail on timeout.

[Nic Brough -Adaptavist-]

No, only the optimisation fails, the index is not damaged. You need to put the defaults back and stop trying to fix the wrong problem.

[na1s]

Hi Nic, I've found in source code of Jira 6.0.3, that Jira doesn't use these params, so they are useless, I tried to return index optimizing job, and it got speed improvement about 3-4 times, but we can't use it in production, because optimizing job locks index for reading and for your system it is not possible, so I don't know how to optimize index:-) Maybe I should run expungeDeletes on IndexWriter, what do you think about it?)

[Nic Brough -Adaptavist-]

I'd simply move on from this. Index optimisation does NOT lock Jira, it just makes ti queue up more index writes for a few seconds. Leave it alone, stop messing with it. Or, explain the actual problem that I'm missing - what you've described above simply is not the way Jira does things, you shouldn't need to touch the indexing at all.