Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.5
-
None
Description
https://buildbot.askmonty.org/buildbot/reports/cross_reference with test "thread_pool_info" shows a lot of failures, mostly in 10.5
Example:
https://buildbot.askmonty.org/buildbot/builders/kvm-deb-bionic-amd64/builds/2721
main.thread_pool_info w1 [ fail ]
Test ended at 2019-08-16 10:29:34
CURRENT_TEST: main.thread_pool_info
— /usr/share/mysql/mysql-test/main/thread_pool_info.result 2019-08-16 07:40:14.000000000 -0400
+++ /dev/shm/var/1/log/thread_pool_info.reject 2019-08-16 10:29:34.265336935 -0400
@@ -55,7 +55,7 @@
FLUSH THREAD_POOL_STATS;
SELECT SUM(DEQUEUES_BY_LISTENER+DEQUEUES_BY_WORKER) FROM INFORMATION_SCHEMA.THREAD_POOL_STATS;
SUM(DEQUEUES_BY_LISTENER+DEQUEUES_BY_WORKER)
-1
+0
SELECT SUM(POLLS_BY_LISTENER+POLLS_BY_WORKER) FROM INFORMATION_SCHEMA.THREAD_POOL_STATS;
SUM(POLLS_BY_LISTENER+POLLS_BY_WORKER)
2
Attachments
Activity
otto cross reference is not available at this point, so that cannot be easily achieved.
I did not find the page that is an overview of branch 10.5, but via a recent commit I found this view: https://buildbot.mariadb.org/#/changes/52636
It confirms that the issue I reported on May 15th does not exists anymore and the builder passes.
After that there was multiple fix commits to threadpool, so seems one of those fixed this issue for good and it is OK to close this bug report.
$ git log --format="%Cgreen%ci%Creset %aE %s" --since=2020-05-01 sql/threadpool*
|
2020-05-25 14:54:11 +0200 wlad@mariadb.com MDEV-22696 TP_pool_generic::set_pool_size logic so that it marks each connection to change group before the next socket read.
|
2020-05-25 14:54:11 +0200 wlad@mariadb.com MDEV-22696 Threadpool : make sure thd->event_scheduler.data does not change as long as THD is in server_threads.
|
2020-05-25 14:54:00 +0200 wlad@mariadb.com Threadpool - support changing group on Windows with generic thread pool
|
2020-05-23 12:29:10 +0300 monty@mariadb.org Change THD->transaction to a pointer to enable multiple transactions
|
2020-05-15 16:29:06 +0200 wlad@mariadb.com MDEV-22578 thread_pool_info crashes with clang6, using SSE instructions on unaligned memory
|
"MDEV-22578 thread_pool_info crashes with clang6, using SSE instructions on unaligned memory" is fix for a regression that was introduced in MDEV-5205 many moons ago , which resulted in a very much reproducible crash with optimized build on clang6.
It is possible that declared alignment on a structure, which is not given due to my_malloc() could cause other problems if compiler takes alignment seriously.
Nothing else has even a remote relation with that bug, and are small performance tweaks, some windows fix, and something monty needed. Impressive detective work though.
Looking at recent Travis-CI jobs at https://travis-ci.org/github/MariaDB/server/branches I did not see any jobs failing on this test, so I hereby confirm that this issue is fixed.
Thanks!
Maybe we should stop abusing the bug report, and somebody just creates a good documentation for the experimental buildbot that is run by the foundation, and announces the availability of the documentation on the mailing list.