ignite-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From neerajbhatt <neerajbhatt2...@gmail.com>
Subject wrong data from cluster nodes while using left join
Date Thu, 13 Jul 2017 11:57:03 GMT
Hi All

We have created cluster groups for each node of a cluster (so that we can
give time out for each node (cluster group), in case query fails in a node
we can get a partial result )

We have couple of partitioned caches (with backup as 1) and we are sending a
query with left join
id is available in all the caches R, I , C and D and there are other filter
criteria on other fields of R , I , C and D

SELECT distinct r.id , r.rank  FROM A.R as r  
LEFT JOIN  B.I as T ON r.id = T.id
LEFT JOIN C.C as cat ON cat.id = T.id
LEFT JOIN D.D as img  ON img.id = T.id
WHERE cat.categoryId != '1' AND  T.itemType = '1'  AND T.rating= '0' 
AND T.reviews='0' AND T.downloads='0' AND  img.imageId='1'  AND
order by rank desc limit 3

This query returns  fast ( around 10 ms) but it is giving results ( ids) 
from cluster group B (node B) too when none of the caches in B have those
ids. The log file in each node (cluster group shows same ids)

If we just use where clause instead of left join the query performance is
decreased significantly (10 second)

Is it because of left join? but in case of left join too when data is not in
caches for a particular node, how the data is being fetched for those ids

View this message in context: http://apache-ignite-users.70518.x6.nabble.com/wrong-data-from-cluster-nodes-while-using-left-join-tp14782.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.

View raw message