How to Produce a Full Stack Trace for mysqld

Partial Stack Traces in the Error Log

When mysqld / mariadbd crashes, it will write a stack trace in the error log by default. This is because the stack_trace option defaults to ON. With a normal release build, this stack trace in the error log may look something like this:

2019-03-28 23:31:08 0x7ff4dc62d700  InnoDB: Assertion failure in file /home/buildbot/buildbot/build/mariadb-10.2.23/storage/innobase/rem/rem0rec.cc line 574
InnoDB: We intentionally generate a memory trap.
InnoDB: Submit a detailed bug report to https://jira.mariadb.org/
InnoDB: If you get repeated assertion failures or crashes, even
InnoDB: immediately after the mysqld startup, there may be
InnoDB: corruption in the InnoDB tablespace. Please refer to
InnoDB: https://mariadb.com/kb/en/library/innodb-recovery-modes/
InnoDB: about forcing recovery.
190328 23:31:08 [ERROR] mysqld got signal 6 ;
This could be because you hit a bug. It is also possible that this binary
or one of the libraries it was linked against is corrupt, improperly built,
or misconfigured. This error can also be caused by malfunctioning hardware.
 
To report this bug, see https://mariadb.com/kb/en/reporting-bugs
 
We will try our best to scrape up some info that will hopefully help
diagnose the problem, but since we have already crashed, 
something is definitely wrong and this may fail.
 
Server version: 10.2.23-MariaDB-10.2.23+maria~stretch
key_buffer_size=134217728
read_buffer_size=131072
max_used_connections=234
max_threads=752
thread_count=273
It is possible that mysqld could use up to 
key_buffer_size + (read_buffer_size + sort_buffer_size)*max_threads = 1783435 K  bytes of memory
Hope that's ok; if not, decrease some variables in the equation.
 
Thread pointer: 0x7ff4d8001f28
Attempting backtrace. You can use the following information to find out
where mysqld died. If you see no messages after this, something went
terribly wrong...
stack_bottom = 0x7ff4dc62ccc8 thread_stack 0x49000
*** buffer overflow detected ***: /usr/sbin/mysqld terminated
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x70bfb)[0x7ffa09af5bfb]
/lib/x86_64-linux-gnu/libc.so.6(__fortify_fail+0x37)[0x7ffa09b7e437]
/lib/x86_64-linux-gnu/libc.so.6(+0xf7570)[0x7ffa09b7c570]
/lib/x86_64-linux-gnu/libc.so.6(+0xf93aa)[0x7ffa09b7e3aa]
/usr/sbin/mysqld(my_addr_resolve+0xe2)[0x55ca42284922]
/usr/sbin/mysqld(my_print_stacktrace+0x1bb)[0x55ca4226b1eb]
/usr/sbin/mysqld(handle_fatal_signal+0x41d)[0x55ca41d0a01d]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x110e0)[0x7ffa0b4180e0]
/lib/x86_64-linux-gnu/libc.so.6(gsignal+0xcf)[0x7ffa09ab7fff]
/lib/x86_64-linux-gnu/libc.so.6(abort+0x16a)[0x7ffa09ab942a]
/usr/sbin/mysqld(+0x40f971)[0x55ca41ab8971]
/usr/sbin/mysqld(+0x887df6)[0x55ca41f30df6]
/usr/sbin/mysqld(+0x863673)[0x55ca41f0c673]
/usr/sbin/mysqld(+0x96648e)[0x55ca4200f48e]
/usr/sbin/mysqld(+0x89b559)[0x55ca41f44559]
/usr/sbin/mysqld(+0x8a15e4)[0x55ca41f4a5e4]
/usr/sbin/mysqld(+0x8a2187)[0x55ca41f4b187]
/usr/sbin/mysqld(+0x8b1a20)[0x55ca41f5aa20]
/usr/sbin/mysqld(+0x7f5c04)[0x55ca41e9ec04]
/usr/sbin/mysqld(_ZN7handler12ha_write_rowEPh+0x107)[0x55ca41d140d7]
/usr/sbin/mysqld(_Z12write_recordP3THDP5TABLEP12st_copy_info+0x72)[0x55ca41b4b992]
/usr/sbin/mysqld(_Z12mysql_insertP3THDP10TABLE_LISTR4ListI4ItemERS3_IS5_ES6_S6_15enum_duplicatesb+0x1206)[0x55ca41b560f6]
/usr/sbin/mysqld(_Z21mysql_execute_commandP3THD+0x3f68)[0x55ca41b6bee8]
/usr/sbin/mysqld(_Z11mysql_parseP3THDPcjP12Parser_statebb+0x28a)[0x55ca41b70e4a]
/usr/sbin/mysqld(+0x4c864f)[0x55ca41b7164f]
/usr/sbin/mysqld(_Z16dispatch_command19enum_server_commandP3THDPcjbb+0x1a7c)[0x55ca41b737fc]
/usr/sbin/mysqld(_Z10do_commandP3THD+0x176)[0x55ca41b748a6]
/usr/sbin/mysqld(_Z24do_handle_one_connectionP7CONNECT+0x25a)[0x55ca41c3ec0a]
/usr/sbin/mysqld(handle_one_connection+0x3d)[0x55ca41c3ed7d]
/usr/sbin/mysqld(+0xb75791)[0x55ca4221e791]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x74a4)[0x7ffa0b40e4a4]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x3f)[0x7ffa09b6dd0f]

If you plan to report a bug about the problem, then this information can be very useful for MariaDB's developers to track down the root cause. However, notice that some of the function names in the call stack are missing. In some cases, this partial stack trace may not be enough to find out exactly where the problem is.

A full stack trace can only be produced if you have debugging symbols for your mysqld binary.

Obtaining Debugging Symbols for Your mariadbd executable

Debug information is used by debugging tools to produce a meaningful stack trace. Importantly these packages do not replace any executables or any existing production executables or in any way interfere with the way the production server ran before these packages where installed.

If you are obtaining a backtrace for a coredump, you can move the core dump to a difference server that has the identical mariadb-server and debug info packages, and perform the backtrace there with no loss of information.

Installing Debug Info Packages on Linux

On some Linux distributions, you may be able to install debuginfo packages that contain debugging symbols.

Currently, debuginfo packages may not allow the server to print a nice stack trace in the error log. They also allow users to extract full stack traces from core dumps. See MDEV-20738 for more information.

Installing Debug Info Packages with yum/dnf

On RHEL, CentOS, Fedora, and other similar Linux distributions, it is highly recommended to install the relevant RPM package from MariaDB's repository using yum or dnf. Starting with RHEL 8 and Fedora 22, yum has been replaced by dnf, which is the next major version of yum. However, yum commands still work on many systems that use dnf. For example:

sudo yum install MariaDB-server-debuginfo

See Installing MariaDB with yum/dnf: Installing Debug Info Packages with YUM for more information.

Installing Debug Info Packages with zypper

On SLES, OpenSUSE, and other similar Linux distributions, it is highly recommended to install the relevant RPM package from MariaDB's repository using zypper. For example:

sudo zypper install MariaDB-server-debuginfo

See Installing MariaDB with zypper: Installing Debug Info Packages with ZYpp for more information.

Installing Debug Info Packages from MariaDB's Debian or Ubuntu repository

These are for when you already installed MariaDB from a MariaDB mirror.

For Ubuntu and additional repository step is needed:

sudo add-apt-repository 'deb [arch=amd64,arm64,ppc64el,s390x]  https://ftp.osuosl.org/pub/mariadb/repo/10.5/ubuntu focal main/debug'

Adjust 10.5 to the major version you are debugging and focal to the required distribution.

apt-get update && apt-get install -y mariadb-server-core-10.5-dbgsym 

Installing Debug Info Packages packaged by Ubuntu or Debian

If you used the MariaDB versions provided by Debian or Ubuntu see the following links.

For Debian see https://wiki.debian.org/AutomaticDebugPackages

For Ubuntu see https://wiki.ubuntu.com/Debug%20Symbol%20Packages

Installing Debugging Symbols on Windows

Debugging symbols are available to install on Windows.

Installing Debugging Symbols with the MSI Installer on Windows

Debugging symbols can be installed with the MSI installer. Debugging symbols are not installed by default. You must perform a custom installation and explicitly choose to install debugging symbols.

The MSI installer can be downloaded from the MariaDB downloads page.

Installing Debugging Symbols with the ZIP Package on Windows

MariaDB also provides a ZIP package that contains debugging symbols on Windows.

The ZIP package that contains debugging symbol can be downloaded from the MariaDB downloads page.

Enabling Core Dumps

To enable core dumps, see Enabling Core Dumps for details.

Where is the Core File on Linux?

At the bottom of the error log there will be some text about the core location including:

Writing a core file...
Working directory at /var/lib/mariadb
Resource Limits:
Limit                     Soft Limit           Hard Limit           Units     
...
Max core file size        unlimited            unlimited            bytes     
...
Core pattern: |/usr/lib/systemd/systemd-coredump %P %u %g %s %t %c %h

Kernel version: Linux version 6.0.0-0.rc2.19.fc38.x86_64 ([email protected]) (gcc (GCC) 12.2.1 20220819 (Red Hat 12.2.1-1), GNU ld version 2.39-2.fc38) #1 SMP PREEMPT_DYNAMIC Mon Aug 22 12:52:40 UTC 2022

If the was a core limit in the resource limits there may be limited or no core file information.

If the core pattern begins with a {{|}}, then the following is the executable that handled the core file during the crash. The following show a few techniques to access the core depending on the pattern. If another program is used, look at its manual page to see how to get access to the core file.

If a plain core filename is in the "Core pattern" there's a good chance it will be in the Working directory location. It might have a .{process number} suffix on the filename.

Extracting a core file from a container

If you are running mariadb in a container, the locations where the core dump can be generated are limited. Looking at the container log, this will likely be where the error log information is. The "Core pattern" of a Linux system is currently a global fixed value. The consequence is if this core pattern refers to a program, that program isn't likely to be in the container and won't be executed on the crash.

The system wide crash handler can be changed with sysctl kern.core_pattern=core to set this back to a file based crash. With this, the crash should occur in the working directory, normally the /var/lib/mysql data directory of the container volume.

Extracting a core file from systemd-coredump

For systemd-coredump, there is a program coredumpctl to manage access.

coredumpctl list
TIME                            PID   UID   GID SIG     COREFILE EXE                                                                                        >
Fri 2022-09-09 14:16:37 AEST 213571  1000  1000 SIGSEGV present  /usr/sbin/mariadbd 

To access the program using gdb, coredumpctl debug (defaults to last crash), will load the core dump in gdb. The instructions in the next section for extracting information.

Extract a core file from abrt

A core pattern of |/usr/libexec/abrt-hook-ccpp indicates abrt system is used.

abrt-cli is a command line user interface for access the core file.

Extract a core file from apport

A core pattern of [|/usr/share/apport/apport indicates apport.

For more information see Apport Project Wiki.

apport-retrace allows you to "Examine Locally" and run a gdb session. One you have gdb started instructions in the next section can be used for extracting information.

Analyzing a Core File with gdb on Linux

To analyze the core file on Linux, you can use gdb.

For example, to open a core file with gdb, you could execute the following:

sudo gdb /usr/sbin/mysqld  /var/lib/mysql/core.932

Be sure to replace /usr/sbin/mysqld with the path to your mysqld binary and to also replace /var/lib/mysql/core.932 with the path to your core file.

Once gdb has opened the core file, if you want to log all output to a file, then you could execute the following commands:

set logging file /tmp/gdb_output.log
set logging on

If you do not execute set logging file, then the set logging on command creates a gdb.txt in your current working directory. Redirecting the output to a file is useful, because it can make it easier to analyze. It also makes it easier to send the information to a MariaDB developer, if that becomes necessary.

Do any commands that you would like to do. For example, you could get the backtraces.

Once you are done, you can exit gdb by executing the quit command.

Getting Backtraces with gdb on Linux

On Linux, once you have debugging symbols for your mysqld binary, you can use the gdb utility to get backtraces, which are what gdb calls stack traces. Backtraces can be obtained from a core file or from a running mysqld process.

Full backtraces are preferred and will contain function arguments, which can contain useful information such as query strings, so it can make the information easier to analyze.

To get a full backtrace of the main thread, then you could execute the following:

bt -frame-arguments all full

If you want to get a full backtrace of all threads, then you could execute the following:

thread apply all bt -frame-arguments all full

If you want to get a full backtrace to a file to report a bug, the recommended way is to use gdb:

set logging on
set pagination off
set print frame-arguments all
thread apply all bt full
set logging off

This will write the full backtrace into the file gdb.txt.

Getting Full Backtraces For All Threads From a Core File

Sometimes it can be helpful to get full backtraces for all threads. The full backtraces will contain function arguments, which can contain useful information such as query strings, so it can make the information easier to analyze.

To get full backtraces for all threads from a mysqld core file, execute a command like the following:

sudo gdb --batch --eval-command="thread apply all bt  -frame-arguments all full" /usr/sbin/mysqld /var/lib/mysql/core.932  > mysqld_full_bt_all_threads.txt

Be sure to replace /usr/sbin/mysqld with the path to your mysqld binary and to also replace /var/lib/mysql/core.932 with the path to your core dump.

The backtraces will be output to the file mysqld_full_bt_all_threads.txt.

Getting Full Backtraces For All Threads From a Running mysqld Process

Sometimes it can be helpful to get full backtraces for all threads. The full backtraces will contain function arguments, which can contain useful information such as query strings, so it can make the information easier to analyze.

To get full backtraces for all threads from a running mysqld process, execute a command like the following:

sudo gdb --batch --eval-command="thread apply all bt  -frame-arguments all full" /usr/sbin/mysqld $(pgrep -xn mysqld)  > mysqld_full_bt_all_threads.txt

Be sure to replace /usr/sbin/mysqld with the path to your mysqld binary.

The backtraces will be output to the file mysqld_full_bt_all_threads.txt.

Running a Copy of the Database Directory

If you are concerned with debuggers running on your production database you can also copy the database to another location.

This is useful when you know which statement crashed the server.

Just start mysqld with the options --datadir=/copy-of-original-data-directory --core-file --stack-trace --socket=/tmp/mysqld-alone.sock --skip-networking

Disabling Stack Traces in the Error Log

In order to disable stack traces in the error log, you can configure the skip_stack_trace option either on the command-line or in a relevant server option group in an option file. For example:

[mariadb]
...
skip_stack_trace

Reporting the Problem

If you encounter some problem in MariaDB, then MariaDB's developers would appreciate if you would report a bug at the MariaDB JIRA bug tracker. Please include the following information:

• Your full stack trace.
• Your error log.
• Your option files.
• How to reproduce the problem.
• SHOW CREATE TABLE {table (for each table in query) and EXPLAIN {query} if a query related crash.

For very difficult or critical errors, you should consider uploading the following information to the MariaDB FTP server in a .tar.gz or .zip archive:

• Your debug build of mysqld (if you compiled it), otherwise version information on the mariadb-server package.
• Your core file.
• Your contact information.
• The associated JIRA issue identifier for the bug, if you reported a bug.

This information will allow the MariaDB developers at the MariaDB Corporation to analyze it and try to create a fix.