[MXS-3863] Maxscale memory leak (2.4.15) Created: 2021-11-10  Updated: 2021-11-11  Resolved: 2021-11-11

Status: Closed
Project: MariaDB MaxScale
Component/s: Core
Affects Version/s: 2.4.15
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Kevin Assignee: Unassigned
Resolution: Not a Bug Votes: 0
Labels: Memory_leak
Environment:

Red Hat Enterprise Linux

Attachments: JPEG File 16365160973721.jpg     JPEG File 16365161868707.jpg    

 Description   

Hi,
Our MaxScale server runs about 200M memory generally, but sometimes it surge to about 20G within 5 minutes, and memory never goes down even after workload peak.

We checked the time of the problem, and we found that a select query was reading a large result set (tens of millions of rows). But when we execute the select query alone, problem cannot be reproduced.

 Comments   
Comment by markus makela [ 2021-11-10 ]

Do you have writeq_high_water configured?

Comment by Kevin [ 2021-11-10 ]

No.
I'll set the parameters to writeq_high_water=32Mi, writeq_low_water=16Mi and have a look.

However I'm still wondering why the memory won't go down but kept staying at that high level.

Comment by markus makela [ 2021-11-10 ]

That is unexpected as that big of a chunk of memory should be returned back to the operating system if it is no longer used. Smaller chunks of memory might be kept around by the memory allocator but large chunks are usually freed.

Comment by Kevin [ 2021-11-11 ]

@markus makela The problem seems to be solved. Thanks a lot for your support!

Comment by markus makela [ 2021-11-11 ]

OK, I'll close this issue as Not a Bug.

Generated at Thu Feb 08 04:24:30 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.