Hi all, Can anyone here explain

/druid/historical/v1/readiness

, a service status api in historical? I was expecting to receive 503 when historical is downloading the segments from deep storage. I have 3 data nodes, and I replaced one of them. The UI shows

xxxx segemnts to load xxxGB

but when I check the response using cURL in the historical node. It returns 200.

Hi Youngsol, It might mean that at least one copy of each segment is loaded ... which in the case of your single node replacement operation, segments should have remained fully available at all times.

I suspect that too, but it does not make sense on kubernetes environemnt. During the rolling restarts, It will return 200 ok due to other historicals that have not been rolled out yet.

Unless I'm wrong, historical are subject of segment loading all the time, whenever coordinator affects some segment to them. If it would rely on that, you would have constant replacement of pods due to readiness not being ok. I don't have the code at the moment but if the historical is up&running, joined the cluster property and started its duty, then it's ready. Whether the segments are loaded or not does not matter. When you issue your query, druid will itself contact the proper historical to respond.

my hack is: i know it takes about 300 seconds, so I set this a startUpProbes with initialDelaySeconds: 300 but the path is

/druid/historical/v1/loadstatus

not

/status/health