Change in rules in Coordinator


My setup:

One node : Hot

Three nodes: _default tier

I did setup one of the datasources to be in hot tier like:


jenkins-events: P1D to be in Hot tier

This configuration worked for sometime and later it didnt do any data collection

I tried restarting realtime node, but didnt work.

I tried to update the period to P3D to be in Hot tier, didnt do anything

No data flow seen to data sources instead.

Do we need to restart any nodes when we change coordinator rules for datasources?

Do we need to configure 2 rules necessarily, if hot else default tier that required to be configured for each datasource?


It’s unclear to me whether the issue is with load rules or with your realtime ingestion. I would recommend first removing all load rules and just leaving the default (load forever, 2 replicas, in _default_tier). If you starting seeing data coming in, then it sounds like it’s rule related, otherwise I would get the ingestion part stable before changing the load rules.

One thing worth noting here is that load rules determine whether a segment gets marked used or unused by the coordinator in the ‘rule matching phase’ of the run; what it does not do is go back and change the used status of previous segments. So if you previously had a week of data in jenkins-events, and you set it to something like “load P1D in Hot tier” followed by “drop forever”, it would mark everything older than a day as unused. If you then changed it to “load P3D in Hot tier”, you would not see the previous 3 days of data come back. In order to apply the rule to segments previously marked unused, you have to ‘enable’ the datasource, either through the UI or through the coordinator API call (POST /druid/coordinator/v1/datasources/{dataSourceName}), as described here:

You don’t need to restart nodes when you change coordinator rules, just re-enable the datasource if you expect some segments previously marked as unused to be enabled again.

You can have as many or few rules as you like. The way the rule logic works is it goes down the list and applies the first rule that matches the period or interval criteria (and does this independently for each tier)

Hope this helps,