This message was deleted.
# troubleshootings
Slackbot
02/06/2023, 4:39 AMThis message was deleted.
a
Adarsh Sanjeev
02/06/2023, 4:51 AMI believe according to the documentation here: https://druid.apache.org/docs/latest/querying/query-context.html#vectorization-parameters
Vectorization does not support joins. It can be enabled with a context param but it is not recommended for uses other than testing.
v
Vijay Narayanan
02/06/2023, 6:13 AMyou could use vectorize=“force”. If the query is not vectorizable then this context will throw an error and you could use that to see which of your queries need to be re-written
👍 1
d
D K
02/06/2023, 10:58 AMalso below are the additional query context params that i am using to improve performance. Please let me know if these params are good enough
D K
02/06/2023, 10:58 AM"useApproximateCountDistinct": true,
"useApproximateTopN": true,
"populateCache": false,
"useCache": false,
"vectorize": true,
"vectorSize": 256,
"forceLimitPushDown": false,
"enableParallelMerge": true
v
Vijay Narayanan
02/06/2023, 11:06 AMenableParallelMerge is true by default no need to set it. forceLimitPushDown is false by default no need to set it. vectorize is also true by default. Changing the default vectorsize from 512 to 256 may not change performance by much
Vijay Narayanan
02/06/2023, 11:10 AMuseApproximateTopN is also true by default. As far as caching is concerned…(“populateCache”: false,
“useCache”: false) this is much trickier. Historical cache per segment results. What this means is that if the exact same query is run and you have both new and old segments the results for the old segments are picked up from cache while results for new segments are computed and the two are combined. This can speed up performance by an order of magnitude when you have the same query running repeatedly on additional data
d
D K
02/06/2023, 12:31 PMthank you very much!
2 Views