[MDEV-13635] MariaDB Server 10.2.6 hangs up during shutdown Created: 2017-08-24  Updated: 2019-10-03

Status: Open
Project: MariaDB Server
Component/s: None
Affects Version/s: 10.2.6
Fix Version/s: 10.2

Type: Bug Priority: Major
Reporter: Michael Graf Assignee: Oleksandr Byelkin
Resolution: Unresolved Votes: 0
Labels: None
Environment:

Windows Server 2016 Datacenter
RAM: 32 GB


Attachments: File HH-VMESX-ZIS01.err     File my.ini     Zip Archive mysqld.exe_170824_104620.zip     Text File show globabl variables.txt     Text File uniqstack_0d5c_2017-08-24_14-07-38-215.txt    
Issue Links:
Duplicate
duplicates MDEV-689 LP:619666 - Deadlock with Aria around... Closed
duplicates MDEV-9722 mysqld.exe hangs after upgrade Closed
Relates
relates to MDEV-13736 MariaDB Server 5.5.54 hangs up during... Stalled

 Description   

MariaDB was shutdown by service manager and hanged up for several hours.
Nothing happend anymore.
The process was killed after 10 hours with the task manager.
A minidump was created before the process was killed.



 Comments   
Comment by Vladislav Vaintroub [ 2017-08-24 ]

uniqstack_0d5c_2017-08-24_14-07-38-215.txt Attached unique callstacks. looks like deadlock in aria

Comment by Vladislav Vaintroub [ 2017-08-24 ]

Looks like MDEV-9722

Comment by Vladislav Vaintroub [ 2017-08-24 ]

2 threads are hanging in translog_flush_wait_for_end() waiting for some condition variable.
Other threads are waiting for some mutex in either open or close of aria tables.
nothing else seems to happen.

Comment by Vladislav Vaintroub [ 2017-08-24 ]

more from the log

{mysqld!log_descriptor
   +0x000 pagecache        : 0x00007ff6`a7502940 st_pagecache
   +0x008 flags            : 0
   +0x00c open_flags       : 0x8002
   +0x010 log_file_max_size : 0x40000000
   +0x014 server_version   : 0x1876e
   +0x018 server_id        : 1
   +0x01c buffer_capacity_chunk_2 : 0xffc00
   +0x020 half_buffer_capacity_chunk_2 : 0x7fe00
   +0x024 page_overhead    : 7
   +0x026 page_capacity_chunk_2 : 0x1ff8
   +0x028 directory        : [512]  "d:\leutek\mariadb\data\"
   +0x228 open_files       : st_dynamic_array
   +0x248 max_file         : 8
   +0x24c min_file         : 8
   +0x250 open_files_lock  : st_mysql_rwlock
   +0x298 directory_fd     : 0n-1
   +0x2a0 buffers          : [8] st_translog_buffer
   +0x800860 dirty_buffer_mask : 0x10 ''
   +0x800868 dirty_buffer_mask_lock : st_mysql_mutex
   +0x800898 horizon          : 0n34812902403
   +0x8008a0 bc               : st_buffer_cursor
   +0x8008c0 max_lsn          : 0n34812902400
   +0x8008c8 flushed          : 0n34812901935
   +0x8008d0 sent_to_disk     : 0n34812902400
   +0x8008d8 log_start        : 0n34713202404
   +0x8008e0 previous_flush_horizon : 0n34812900802
   +0x8008e8 in_buffers_only  : 0n34812895232
   +0x8008f0 sent_to_disk_lock : st_mysql_mutex
   +0x800920 log_flush_lock   : st_mysql_mutex
   +0x800950 log_flush_cond   : st_mysql_cond
   +0x800960 new_goal_cond    : st_mysql_cond
   +0x800970 file_header_lock : st_mysql_mutex
   +0x8009a0 unfinished_files_lock : st_mysql_mutex
   +0x8009d0 unfinished_files : st_dynamic_array
   +0x8009f0 min_need_file    : 8
   +0x8009f4 min_file_number  : 8
   +0x8009f8 purger_lock      : st_mysql_mutex
   +0x800a28 last_lsn_checked : 0n0
   +0x800a30 is_everything_flushed : 1 ''
   +0x800a31 flush_in_progress : 0 ''
   +0x800a34 flush_no         : 0n436282
   +0x800a38 next_pass_max_lsn : 0n0
   +0x800a40 max_lsn_requester : 0x1df4

The thread that is max_lsn_requester is actually currently blocking in the mutex in ma_open(). sanja says the flush is not ongoing anymore. However 2 threads are waiting for the end of flush.

Comment by Michael Graf [ 2017-09-13 ]

The same problem occurred with MariaDB 5.5.54 - see MDEV-13736[#https://jira.mariadb.org/browse/MDEV-13736].
Do you need further help from us ?

Generated at Thu Feb 08 08:07:08 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.