[MXS-403] Monitor callback to DCBs evades thread control causing crashes Created: 2015-10-11  Updated: 2015-11-27  Resolved: 2015-11-27

Status: Closed
Project: MariaDB MaxScale
Component/s: Core, galeramon, mariadbmon, ndbclustermon
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:

Any



 Description   

DCB closure is managed by the zombie process and covers all the threads that are working with the poll list. However, monitors are outside this group of threads, and acting on a DCB through the callback mechanism evades the thread control in the zombie process. This can result in a crash through a DCB being prematurely freed.

The preferred solution is to use the existing mechanism for inserting a "fake" event into the polling so that the change of state (no longer responding) can be produced via the normal poll handling mechanism.



 Comments   
Comment by martin brampton (Inactive) [ 2015-11-27 ]

Fixed in branch MXS-329, now merged into develop for inclusion in next release i.e. 1.3.0. Changed back the title which was no longer English.

Generated at Thu Feb 08 03:59:01 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.