Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.1.4
-
Enterprise Edition 7.1.4 build 3630
-
Untriaged
-
Linux x86_64
-
-
0
-
No
-
Analytics Sprint 20
Description
While running capella scaling tests, it is observed that some of the analytics queries are not honouring the timeout values and are running for a long time.
Steps to reproduce on capella -
- Create a 12 node cluster with 3 KV, 3 N1QL, 3 Index and 3 CBAS node.
- Create a KV bucket and create 2 collections under _default scope.
- Load 1 million docs in each collection.
- Create indexes on collection
- Create a total of 10 datasets on either of the collections.
- Create indexes on the datasets.
- Start parallel analytics queries. For this instance of test 10 query thread are executed parallely from the automation script.
- Start scaling of cluster by changing the instance type of nodes.
- Automation uses Java SDK client to execute analytics queries. The default timeout for analytics queries is 75 sec.
- It is observed that some queries honour this timeout value and gets cancelled, while some do not and run for a long periods.
Indexes created on datasets -
HotelIndexes = [
|
"create index {0} on {1}(phone:string)", |
"create index {0} on {1}(`type`:string)", |
"create index {0} on {1}(city:string, country:string)", |
"create index {0} on {1}(price:double)", |
"create index {0} on {1}(avg_rating:double)", |
]
|
Analytics queries that are executed -
HotelQueries = [
|
"SELECT * from {} where phone like \"4%\" limit 100", |
|
|
"SELECT * from {} where `type` is not null limit 100", |
|
|
"SELECT COUNT(*) FILTER (WHERE free_breakfast = TRUE) AS count_free_breakfast, COUNT(*) FILTER (WHERE free_parking = TRUE) AS count_free_parking, COUNT(*) FILTER (WHERE free_breakfast = TRUE AND free_parking = TRUE) AS count_free_parking_and_breakfast FROM {} WHERE city LIKE 'North%' ORDER BY count_free_parking_and_breakfast DESC limit 100", |
|
|
"WITH city_avg AS (SELECT city, AVG(price) AS avgprice FROM {0} WHERE price IS NOT NULL GROUP BY city) SELECT h.name, h.price FROM {0} h JOIN city_avg ON h.city = city_avg.city WHERE h.price < city_avg.avgprice AND h.price IS NOT NULL limit 100", |
|
|
"select city,country,count(*) from {} where free_breakfast=True and free_parking=True group by country,city order by country,city limit 100", |
|
|
"select avg(price) as AvgPrice, min(price) as MinPrice, max(price) as MaxPrice from {} where free_breakfast=True and free_parking=True and price is not null and array_count(public_likes)>5 and `type`='Hotel' group by country limit 100", |
]
|
|
|
{} - replace this with any dataset names. Dataset names in the logs with be of format ds0, ds1 .... ds9 |
Attachments
Issue Links
- is triggering
-
-
- Closed
-
- links to