Bug #79457 Memory consumtion and request errors during table scan
Submitted: 30 Nov 2015 13:51 Modified: 19 Feb 17:20
Reporter: Mike Pentagra Email Updates:
Status: Need Feedback Impact on me:
None 
Category:MySQL Cluster: NDB API Severity:S5 (Performance)
Version:7.4.7 OS:Oracle Linux
Assigned to: Bogdan Kecman CPU Architecture:Any

[30 Nov 2015 13:51] Mike Pentagra
Description:
We observe high memory consumption and, as a result of swapping due to high memory load, request errors on client side while doing table scan and parallel requests to another table within same transaction.
Table size - about 80k records.

How to repeat:
Scan table using NDB api, for each received data from scaned table request data by id from another table within one transaction.
Sample tables and code in attachments.
[30 Nov 2015 13:53] Mike Pentagra
sample tables and some to reporoduce the problem

Attachment: tables.tar.bz2 (application/x-bzip2, text), 1.72 MiB.

[30 Nov 2015 13:53] Mike Pentagra
sample tables and some to reporoduce the problem

Attachment: ndb_ty.cpp (application/octet-stream, text), 5.67 KiB.

[19 Feb 17:20] Bogdan Kecman
Hi,

I need some additional details from you.

* How did you configure your cluster?
* Where do you see increased data usage, on data nodes or on the api node?

thanks