Query TimeOut on Broker and historical

Hi Guys

I am using Druid-0.7.0.

Query to broker - some log found in broker and historical node

2015-04-28T12:05:41,740 INFO [qtp2054296265-161] io.druid.server.QueryResource - Query timeout [a8dcbefc-da06-4cdc-8dd6-32630d3ad9d2]

My broker is 15 GB and 4 cpu:

Hi Jitesh,

query timeouts are generally caused by wrong configuration

  1. modify druid.processing.numThreads to numCores - 1 i.e 7.
  2. what is the total amount of data loaded and how many historical nodes do you have ? your maxSize is set to 200G, decrease that so that all your data can be loaded in memory.
  3. If you are not running in memory, do you have SSD storage on historicals ? when paging data from disk using SSDs gives much better performance than sata disks.
  4. what are the jvm configs of both broker and historical ? do you see any gc issues on the either the broker or historical node ?
  5. what is the size of individual segments in the cluster ? having very large segments can result in high segment scan time and very small segments increase merging overhead on the broker we try to keep segment sizes nearly 500M per segment.

you will also find http://druid.io/docs/latest/Performance-FAQ.html helpful in tuning the configs.