1. LIMIT make no sense without ORDER for non-local queries and when sqlQueryParallelizm > 1. As in that case, data from different index segments will be merged out of order and you can get different results with same query from run to run .
2. You use ClusterGroup to make your jobs to be run on certain nodes, but it doesn't affect operations inside a job. So, non-local sql query will be run on all data nodes. On Mon, Jul 17, 2017 at 12:44 PM, neerajbhatt <[email protected]> wrote: > Hi Andrew > > It seems to be an order by issue. We have created a test project in github > https://github.com/neerajbhatt/testIgnite > > If we use PutItems to put some records in a partitioned cache(TESTCACHE) > args[0]=number of records, and then Get the items by > GetItemsWithClusterGroup args[0] = query type (0 or 1), args[1]= number of > queries and args [3] as last digit of ip. > > for first query ("SELECT T.id FROM " + "TESTCache.TestItem as T LIMIT 3") > different ids are given for different cluster groups( each cluster group > having one node) > > for second query ("SELECT T.id FROM " + "TESTCache.TestItem as T order by > field1 desc LIMIT 3") we are getting same ids for each cluster group. It > seems in case of order by instead of cluster group data is being fetched > from whole cluster. > > Please suggest > > > > -- > View this message in context: http://apache-ignite-users. > 70518.x6.nabble.com/sql-query-in-case-of-cluster-group-tp14884p14994.html > Sent from the Apache Ignite Users mailing list archive at Nabble.com. > -- Best regards, Andrey V. Mashenkov
