[Maria-discuss] can not kill process
hi i have a strange effect a simple insert got stuck (why ?)
i tried to kill but no success, it is even marked as "killed" (see below)
1398867 kpc localhost kpc 1716 Killed INSERT INTO
i can kill other processes-
my system: 10.5.15-MariaDB engine aria 5.10.0-14-amd64
any ideas what may the problem ? (no usefull entries in any log, as you expected )
re,
Am 11.03.23 um 10:53 schrieb Radisson97@gmx.de:
hi i have a strange effect a simple insert got stuck (why ?)
i tried to kill but no success, it is even marked as "killed" (see below)
1398867 kpc localhost kpc 1716 Killed INSERT INTO
i can kill other processes-
my system: 10.5.15-MariaDB engine aria 5.10.0-14-amd64
any ideas what may the problem ? (no usefull entries in any log, as you expected )
re,
i forgot to mention what i did to "solve" the problem. In the end i used shutdown the database, it also got stuck and only a kill -9 could resolv the situation.
The start was ok, no special problems.
re
Hi,
On Sun, Mar 12, 2023 at 12:45 PM Radisson Radisson97@gmx.de wrote:
Am 11.03.23 um 10:53 schrieb Radisson97@gmx.de:
hi i have a strange effect a simple insert got stuck (why ?)
I am afraid that the question could only be answered if you had attached a debugger to the mariadbd process and produced a stack trace of all threads. In https://mariadb.com/kb/en/how-to-produce-a-full-stack-trace-for-mysqld/ you can read how this is done. Installing the -debuginfo or -dbgsym package for the server makes the stack traces much more useful.
any ideas what may the problem ?
We have observed the InnoDB hang https://jira.mariadb.org/browse/MDEV-29835 on MariaDB 10.6 and later versions in our internal testing. I believe that this type of a hang is possible starting with MariaDB 10.2 and MySQL 5.7.
Furthermore, I suspect that the InnoDB "os_event" that was removed in MariaDB Server 10.6 could occasionally lose signals. We occasionally observe test failures on our CI systems due to that.
It is hard to tell which type of hang would be more likely, and your hang could also have occurred due to something completely different.
Marko
Am 13.03.23 um 07:50 schrieb Marko Mäkelä:
Hi,
On Sun, Mar 12, 2023 at 12:45 PM Radisson Radisson97@gmx.de wrote:
Am 11.03.23 um 10:53 schrieb Radisson97@gmx.de:
hi i have a strange effect a simple insert got stuck (why ?)
I am afraid that the question could only be answered if you had attached a debugger to the mariadbd process and produced a stack trace of all threads. In https://mariadb.com/kb/en/how-to-produce-a-full-stack-trace-for-mysqld/ you can read how this is done. Installing the -debuginfo or -dbgsym package for the server makes the stack traces much more useful.
thx for the info, if it happens again and if possible i will try gdb to see where its stuck. Its a production machine, so do not hold your breath :)
CU
any ideas what may the problem ?
We have observed the InnoDB hang https://jira.mariadb.org/browse/MDEV-29835 on MariaDB 10.6 and later versions in our internal testing. I believe that this type of a hang is possible starting with MariaDB 10.2 and MySQL 5.7.
Furthermore, I suspect that the InnoDB "os_event" that was removed in MariaDB Server 10.6 could occasionally lose signals. We occasionally observe test failures on our CI systems due to that.
It is hard to tell which type of hang would be more likely, and your hang could also have occurred due to something completely different.
Marko
On Thu, Mar 16, 2023 at 11:10 PM Radisson Radisson97@gmx.de wrote:
if it happens again and if possible i will try gdb to see where its stuck. Its a production machine, so do not hold your breath :)
You can enable core dumps, so if the built-in InnoDB watchdog triggers an abort of the server process (a lock wait time exceeds innodb_fatal_semaphore_wait_threshold), you should get a core dump of the crash. Starting with 10.3, the core dump does not include some large buffers, such as the InnoDB buffer pool, so hopefully it should not be unacceptably large.
Yesterday, I pushed the fix of https://jira.mariadb.org/browse/MDEV-29835 (a hang that was introduced in MySQL 5.7 and MariaDB 10.2.2) to the 10.6 branch. Today, I have pushed a merge of it to all 10.x branches. The 11.0 and 11.1 branches should follow within an hour.
The next scheduled quarterly releases should be around April 27.
Marko
participants (3)
-
Marko Mäkelä -
Radisson -
Radisson97＠gmx.de