[MXS-408] Connections to backend databases do not clear promptly Created: 2015-10-15  Updated: 2015-10-18  Resolved: 2015-10-18

Status: Closed
Project: MariaDB MaxScale
Component/s: Core
Affects Version/s: 1.2.1
Fix Version/s: 1.3.0

Type: Bug Priority: Major
Reporter: martin brampton (Inactive) Assignee: martin brampton (Inactive)
Resolution: Fixed Votes: 0
Labels: None
Environment:

Tested CentOS 6.5 but probably all



 Description   

Running a test that consists of constantly repeated runs of a PHP program that makes a database connection, runs a simple SQL query, disconnects causes a large number of processes to appear on the backend databases. Backend configuration single master plus single slave. MaxScale using read-write router or connection router and 6 threads. The effect is seen on both master and slave.

Running the same test directly against the backend master server shows one or zero connections in the process list.

Running the test with MaxScale configured for a single thread eliminates the problem, suggesting that the zombie processing is not clearing quickly. The effect is also much smaller if 2 threads are configured. Enabling or disabling persistent connections has relatively little impact.

Hardware environment is a Digital Ocean VPS with a single core.



 Comments   
Comment by martin brampton (Inactive) [ 2015-10-18 ]

Appears to be fixed in MXS-329 branch.

Generated at Thu Feb 08 03:59:03 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.