Hi Team, I wanted to know the meaning of

query/wait/time

metrics in historical wrt to query ID AND datasource. Does it mean query is waiting for a thread being provided to read the segment or so? I see at heavy load 70% of

query/time

is

query/wait/time

. What should I do improve and reduce the wait time

that is correct, this means there are not enough processing threads to handle the queries. • may be the segment size is not apt and each query is taking too much time to process, hence rest of the queries are just waiting. For this either you need to size the segment properly so that queries can run faster or increase the historical nodes, so that segments can be distributed across the nodes. • Check if how much memory left after providing for historicals like Heap + buffers. Rest of the memory is supposed to be used by page cache so that OS can read the recently accessed data from page cache instead of disk. • Increase the processing threads (Which means you might have to increase the resource as well, as suggest config for processing thread = #core -1 ) https://druid.apache.org/docs/latest/operations/segment-optimization