wojciech.kapcia@tigase.net opened 2 years ago
|
|||||||
Fixes applied. https://github.com/tigase/tigase-pubsub/pull/21 |
|||||||
I've updated -dist project to include new versions of MA & UA (server, pubsub and muc were using SNAPSHOTS already). |
|||||||
Merged, let's wait for the TTS-NG results and backport it if evenrything is OK (though, this one would be good candidate to test repositories with TestContainers). Curiously enough, we had those handlers (without |
|||||||
The tests passed (https://tc.tigase.net/buildConfiguration/TigaseTtsNgTests_MySQL/173170). Could you backport it to tigase-server-8.2.x (it had issue #5.0.0) |
|||||||
Handlers without resignal were catching exceptions and returning to the application like the statement (stored procedure) was executed correctly and that caused invalid state in the application level cache. |
|||||||
Should I backport only fixes for PubSub or all the applicable fixes for version 8.2.x? |
|||||||
I've created a new branch in Is there anything else to be done? I'm wondering if we will need to release 5.1.x bugfix as well.. as 5.0.1 would be superior to 5.1.0 in this area.. but simple merge from 5.0.x will not be sufficient as those changes would not be loaded - files should be renamed to 5.1.1 if I recall correctly. |
|||||||
|
|||||||
I've applied mentioned changes, renamed branch, fixed procedure and backported changes for |
|||||||
Thank you. |
|||||||
Fix resolved the issue |
|||||||
wojciech.kapcia@tigase.net added "Related" Customers/carnow-inc#148 12 months ago
|
|||||||
wojciech.kapcia@tigase.net batch edited 6 months ago
|
Type |
Bug
|
Priority |
Normal
|
Assignee | |
Version |
tigase-server-8.4.0, tigase-server-8.2.2
|
-
tigase-server-8.4.0 Closed
Execution concurrently same stored procedure that has transaction inside leads to lock.
Adding exit handler to stored procedures that have the transactions solves the issue:
@andrzej.wojcik please check remaining components (muc, MA, UA, MIX) for stored procedures that are still used and use transactions and add same fix (and please include the comment to not remove the handler).
It would be ideal to have a test for this case (possibly unit-test, but it would be problematic)
To be backported to Tigase Server 8.2.2
Some context: https://stackoverflow.com/a/60230093/211453:
~
MySQL has a timeout for locks:
innodb_lock_wait_timeout
, which by default is 50s