This message was deleted.
# troubleshootings
Slackbot
02/27/2023, 2:35 PMThis message was deleted.
n
Nick M
02/27/2023, 3:11 PMWhen the coordinator is restarted, it takes some time for it to learn about all the segments in metadata and various other things. Depending on how many segments you have in your metadata, this can last for tens of minutes (up to an hour if you're talking millions of segments).
If you enable debug logging, you can see this happening in the logs...
s
santhosh K
02/27/2023, 4:14 PMIn that mean, will the ingestion and other application queries work proper?
v
Vadim
02/27/2023, 5:47 PMFYI, in Druid 25 there were changes made to the console so it works better when the coordinator is being restarted
🔥 1
s
santhosh K
02/27/2023, 5:51 PMokay, but my point is, will there be any impact during the UI is stuck ?
g
Gian Merlino
03/01/2023, 2:43 AMif the time period is not too long (a few minutes) everything will be 100% fine. if it's very long (tens of minutes/hours) then you may have issues with delayed handoff from ingest tasks, and delayed rebalancing of historicals (in the case where rebalancing is needed).
Gian Merlino
03/01/2023, 2:43 AMwe're working on improving these latter cases for very large clusters with very large numbers of segments; if you're in these cases let us know!
we may be able to suggest fixes to any particular issues you're having, and we're also always interested in feedback from people that run very large clusters
s
santhosh K
04/05/2023, 6:25 AMThanks @Gian Merlino
santhosh K
04/05/2023, 6:26 AM@Gian Merlino, I am stuck with a different issue, where Druid is unaware of HA on deep storage HDFS, when ever we switch namenode, we are doing change in middle managers to change /etc/hosts with name node as current active.. Is there a better way to handle this?
g
Gian Merlino
04/20/2023, 4:40 AMhmm, do you have HA configured in
hdfs-site.xmlGian Merlino
04/20/2023, 4:40 AMif so you could put this in your
_commonGian Merlino
04/20/2023, 4:40 AMIt should get picked up by the extension