[MXS-1875] MaxScale crashes 5 minutes after OS startup Created: 2018-05-18  Updated: 2020-08-25  Resolved: 2018-05-28

Status: Closed
Project: MariaDB MaxScale
Component/s: Core
Affects Version/s: 2.2.5
Fix Version/s: 2.2.7

Type: Bug Priority: Major
Reporter: Valerii Kravchuk Assignee: Johan Wikman
Resolution: Fixed Votes: 1
Labels: None


 Description   

The following backtrace is written to the log:

2018-05-17 15:45:35   alert  : Fatal: MaxScale 2.2.5 received fatal signal 11. Attempting backtrace.
2018-05-17 15:45:35   alert  : Commit ID: d9e69413b4d18da24aa04e68e3ce9208159a267a System name: Linux Release string: CentOS release 6.9 (Final)
2018-05-17 15:45:35   alert  :   /usr/bin/maxscale() [0x406d09]: ??:0
2018-05-17 15:45:35   alert  :   /lib64/libpthread.so.0(+0xf7e0) [0x7fb7be6747e0]: sigaction.c:0
2018-05-17 15:45:35   alert  :   /usr/lib64/maxscale/libmaxscale-common.so.1.0.0(_ZN8maxscale6Worker3getEi+0xa) [0x7fb7beb7c46a]: /home/ec2-user/MaxScale/server/core/worker.cc:538
2018-05-17 15:45:35   alert  :   /usr/lib64/maxscale/libmaxscale-common.so.1.0.0(dcb_add_to_list+0x41) [0x7fb7beb438d1]: /home/ec2-user/MaxScale/server/core/dcb.cc:2799
2018-05-17 15:45:35   alert  :   /usr/lib64/maxscale/libmaxscale-common.so.1.0.0(_ZN8maxscale12MessageQueue18handle_poll_eventsEij+0x137) [0x7fb7beb552a7]: /home/ec2-user/MaxScale/server/core/messagequeue.cc:216
2018-05-17 15:45:35   alert  :   /usr/lib64/maxscale/libmaxscale-common.so.1.0.0(_ZN8maxscale6Worker15poll_waiteventsEv+0x21b) [0x7fb7beb7e21b]: /home/ec2-user/MaxScale/server/core/worker.cc:1207
2018-05-17 15:45:35   alert  :   /usr/lib64/maxscale/libmaxscale-common.so.1.0.0(_ZN8maxscale6Worker3runEv+0x121) [0x7fb7beb7e4f1]: /home/ec2-user/MaxScale/include/maxscale/log_manager.h:116
2018-05-17 15:45:35   alert  :   /usr/bin/maxscale(main+0x1a02) [0x408a72]: /home/ec2-user/MaxScale/server/core/gateway.cc:2254
2018-05-17 15:45:35   alert  :   /lib64/libc.so.6(__libc_start_main+0xfd) [0x7fb7bc75dd1d]: ??:0
2018-05-17 15:45:35   alert  :   /usr/bin/maxscale() [0x4043b9]: ??:0

Crash happens after upgrade from MaxScale 2.1.x.



 Comments   
Comment by Johan Wikman [ 2018-05-21 ]

MaxAdmin connections must be handled by the main thread, so if a connection intended for MaxAdmin arrives on another thread, the connection is moved to the main thread by sending a message using an inter-thread message mechanism.

However, if that connection is closed before the message arrives to the main thread, then the handling of that message in the main thread will cause the crash above.

Generated at Thu Feb 08 04:10:03 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.