This message was deleted.
# general
s
This message was deleted.
p
yes, why you need more than one coordinator and one overlord?
а
I am planning to increase the number of data nodes from 100 to 200 and am wondering whether I need to increase the power/number of the Master node or not. Is there any dependency of the number/power of Master nodes in relation to the number of serviced Data nodes?
p
How long does it take to complete coordinator cycle? What is coordinator's period?
а
4 days
p
is it period or time to complete?
а
time to complete
full restore data from deep storage to all data nods
p
and if data is loaded, how much time coordinator needs to complete coordinations process?
а
about an hour
s
The `coordinator`'s workload grows primarily with the number of segments in the cluster. One hour for a single coordinator pass is way too much. This is meant to run once a minute by default. How many segments do you have? What are their min, avg, max sizes, have you looked at segment size optimization? Are you using compaction? The
overlord
is the second leader process which can run on the same leader node as the coordinator or a different one. The overlord coordinates task execution which affects ingestion, compaction and MSQ workloads. It's workload is determined mostly by the number of tasks it manages.