https://pinot.apache.org/ logo
Join Slack
Powered by
Hi team, I’ve a question about startree index. If ...
# troubleshooting
a
Hi team, I’ve a question about startree index. If I add more columns in dimensionsSplitOrder in startree index config, and restart servers, will the existing segments recreate startree index based on new startree index config? Or just new segments will create startree index based on new startree config?
m
Existing ones will be updated too. You don’t need to restart servers, you can call reload api
a
Got it, thanks
I’ve a table with 100 dimensions, is it a good way to add them all to dimensionsSplitOrder?
m
Not really, that would drastically reduce the scope of preaggregation. You want to split on commonly used filter and group by dimensions, where the query selectivity is low
k
There is a flag in tableconfig to enable this feature to dynamically generate star tree index
m
Yes, 
enableDynamicStarTreeeCreation
need to be set to true in Indexing config
k
You can also try the auto StarTree index config and see if it works well
Don’t try in prod
Try in dev first
a
Both realtime and offline type table could update startree idnex based on updated startree index config by calling reload api?
Hey, I edited startree index config through this ‘edit Table’ button and saved it, then clicked ‘Reload All segments’. When I check server segments size, I didn’t see startree index rebuild and segment size didn’t change. I also tried posting /segments/{tableName}/reload, and didn’t see rebuilding startree index. Then I restarted servers, and rebuilding worked. Am I missing something here?
m
Yes, 
enableDynamicStarTreeeCreation
need to be set to true in Indexing config. Also as Kishore mentioned, please try out in dev cluster first, as star tree index generation is quite resource intensive
a
I did set enableDynamicStarTreeeCreation true.
Good suggestion, I’ll test in dev env first. Thanks.
Open in Slack
PreviousNext
Powered by