mariadb/scripts/mysqld_safe.sh
Anirudh Mangipudi 8977c8fa98 Bug #16776528 RACE CONDITION CAN CAUSE MYSQLD TO REMOVE SOCKET FILE ERRANTLY
Problem Description:
A mysqld_safe instance is started. An InnoDB crash recovery begins which takes
few seconds to complete. During this crash recovery process happening, another
mysqld_safe instance is started with the same server startup parameters. Since
the mysqld's pid file is absent during the crash recovery process the second
instance assumes there is no other process and tries to acquire a lock on the
ibdata files in the datadir.  But this step fails and the 2nd instance keeps 
retrying 100 times each with a delay of 1 second. Now after the 100 attempts, 
the server goes down, but while going down it hits the mysqld_safe script's 
cleanup section and without any check it blindly deletes the socket and pid 
files. Since no lock is placed on the socket file, it gets deleted.

Solution:
We create a mysqld_safe.pid file in the datadir, which protects the presence 
server instance resources by storing the mysqld_safe's process id in it. We
place a check if the mysqld_safe.pid file is existing in the datadir. If yes
then we check if the pid it contains is an active pid or not. If yes again,
then the scripts logs an error saying "A mysqld_safe instance is already 
running". Otherwise it will log the present mysqld_safe's pid into the 
mysqld_safe.pid file.
2013-08-12 21:54:50 +05:30

678 lines
19 KiB
Bash

#!/bin/sh
# Copyright Abandoned 1996 TCX DataKonsult AB & Monty Program KB & Detron HB
# This file is public domain and comes with NO WARRANTY of any kind
#
# Script to start the MySQL daemon and restart it if it dies unexpectedly
#
# This should be executed in the MySQL base directory if you are using a
# binary installation that is not installed in its compile-time default
# location
#
# mysql.server works by first doing a cd to the base directory and from there
# executing mysqld_safe
KILL_MYSQLD=1;
MYSQLD=
niceness=0
# Initial logging status: error log is not open, and not using syslog
logging=init
want_syslog=0
syslog_tag=
user='@MYSQLD_USER@'
pid_file=
err_log=
syslog_tag_mysqld=mysqld
syslog_tag_mysqld_safe=mysqld_safe
trap '' 1 2 3 15 # we shouldn't let anyone kill us
# MySQL-specific environment variable. First off, it's not really a umask,
# it's the desired mode. Second, it follows umask(2), not umask(3) in that
# octal needs to be explicit. Our shell might be a proper sh without printf,
# multiple-base arithmetic, and binary arithmetic, so this will get ugly.
# We reject decimal values to keep things at least half-sane.
umask 007 # fallback
UMASK="${UMASK-0640}"
fmode=`echo "$UMASK" | sed -e 's/[^0246]//g'`
octalp=`echo "$fmode"|cut -c1`
fmlen=`echo "$fmode"|wc -c|sed -e 's/ //g'`
if [ "x$octalp" != "x0" -o "x$UMASK" != "x$fmode" -o "x$fmlen" != "x5" ]
then
fmode=0640
echo "UMASK must be a 3-digit mode with an additional leading 0 to indicate octal." >&2
echo "The first digit will be corrected to 6, the others may be 0, 2, 4, or 6." >&2
fi
fmode=`echo "$fmode"|cut -c3-4`
fmode="6$fmode"
if [ "x$UMASK" != "x0$fmode" ]
then
echo "UMASK corrected from $UMASK to 0$fmode ..."
fi
defaults=
case "$1" in
--no-defaults|--defaults-file=*|--defaults-extra-file=*)
defaults="$1"; shift
;;
esac
usage () {
cat <<EOF
Usage: $0 [OPTIONS]
--no-defaults Don't read the system defaults file
--defaults-file=FILE Use the specified defaults file
--defaults-extra-file=FILE Also use defaults from the specified file
--ledir=DIRECTORY Look for mysqld in the specified directory
--open-files-limit=LIMIT Limit the number of open files
--core-file-size=LIMIT Limit core files to the specified size
--timezone=TZ Set the system timezone
--mysqld=FILE Use the specified file as mysqld
--mysqld-version=VERSION Use "mysqld-VERSION" as mysqld
--nice=NICE Set the scheduling priority of mysqld
--skip-kill-mysqld Don't try to kill stray mysqld processes
--syslog Log messages to syslog with 'logger'
--skip-syslog Log messages to error log (default)
--syslog-tag=TAG Pass -t "mysqld-TAG" to 'logger'
All other options are passed to the mysqld program.
EOF
exit 1
}
my_which ()
{
save_ifs="${IFS-UNSET}"
IFS=:
ret=0
for file
do
for dir in $PATH
do
if [ -f "$dir/$file" ]
then
echo "$dir/$file"
continue 2
fi
done
ret=1 #signal an error
break
done
if [ "$save_ifs" = UNSET ]
then
unset IFS
else
IFS="$save_ifs"
fi
return $ret # Success
}
log_generic () {
priority="$1"
shift
msg="`date +'%y%m%d %H:%M:%S'` mysqld_safe $*"
echo "$msg"
case $logging in
init) ;; # Just echo the message, don't save it anywhere
file) echo "$msg" >> "$err_log" ;;
syslog) logger -t "$syslog_tag_mysqld_safe" -p "$priority" "$*" ;;
*)
echo "Internal program error (non-fatal):" \
" unknown logging method '$logging'" >&2
;;
esac
}
log_error () {
log_generic daemon.error "$@" >&2
}
log_notice () {
log_generic daemon.notice "$@"
}
eval_log_error () {
cmd="$1"
case $logging in
file) cmd="$cmd >> "`shell_quote_string "$err_log"`" 2>&1" ;;
syslog)
# mysqld often prefixes its messages with a timestamp, which is
# redundant when logging to syslog (which adds its own timestamp)
# However, we don't strip the timestamp with sed here, because
# sed buffers output (only GNU sed supports a -u (unbuffered) option)
# which means that messages may not get sent to syslog until the
# mysqld process quits.
cmd="$cmd 2>&1 | logger -t '$syslog_tag_mysqld' -p daemon.error"
;;
*)
echo "Internal program error (non-fatal):" \
" unknown logging method '$logging'" >&2
;;
esac
#echo "Running mysqld: [$cmd]"
eval "$cmd"
}
shell_quote_string() {
# This sed command makes sure that any special chars are quoted,
# so the arg gets passed exactly to the server.
echo "$1" | sed -e 's,\([^a-zA-Z0-9/_.=-]\),\\\1,g'
}
parse_arguments() {
# We only need to pass arguments through to the server if we don't
# handle them here. So, we collect unrecognized options (passed on
# the command line) into the args variable.
pick_args=
if test "$1" = PICK-ARGS-FROM-ARGV
then
pick_args=1
shift
fi
for arg do
val=`echo "$arg" | sed -e "s;--[^=]*=;;"`
case "$arg" in
# these get passed explicitly to mysqld
--basedir=*) MY_BASEDIR_VERSION="$val" ;;
--datadir=*) DATADIR="$val" ;;
--pid-file=*) pid_file="$val" ;;
--user=*) user="$val"; SET_USER=1 ;;
# these might have been set in a [mysqld_safe] section of my.cnf
# they are added to mysqld command line to override settings from my.cnf
--log-error=*) err_log="$val" ;;
--port=*) mysql_tcp_port="$val" ;;
--socket=*) mysql_unix_port="$val" ;;
# mysqld_safe-specific options - must be set in my.cnf ([mysqld_safe])!
--core-file-size=*) core_file_size="$val" ;;
--ledir=*) ledir="$val" ;;
--mysqld=*) MYSQLD="$val" ;;
--mysqld-version=*)
if test -n "$val"
then
MYSQLD="mysqld-$val"
else
MYSQLD="mysqld"
fi
;;
--nice=*) niceness="$val" ;;
--open-files-limit=*) open_files="$val" ;;
--open_files_limit=*) open_files="$val" ;;
--skip-kill-mysqld*) KILL_MYSQLD=0 ;;
--syslog) want_syslog=1 ;;
--skip-syslog) want_syslog=0 ;;
--syslog-tag=*) syslog_tag="$val" ;;
--timezone=*) TZ="$val"; export TZ; ;;
--help) usage ;;
*)
if test -n "$pick_args"
then
append_arg_to_args "$arg"
fi
;;
esac
done
}
#
# First, try to find BASEDIR and ledir (where mysqld is)
#
if echo '@pkgdatadir@' | grep '^@prefix@' > /dev/null
then
relpkgdata=`echo '@pkgdatadir@' | sed -e 's,^@prefix@,,' -e 's,^/,,' -e 's,^,./,'`
else
# pkgdatadir is not relative to prefix
relpkgdata='@pkgdatadir@'
fi
MY_PWD=`pwd`
# Check for the directories we would expect from a binary release install
if test -n "$MY_BASEDIR_VERSION" -a -d "$MY_BASEDIR_VERSION"
then
# BASEDIR is already overridden on command line. Do not re-set.
# Use BASEDIR to discover le.
if test -x "$MY_BASEDIR_VERSION/libexec/mysqld"
then
ledir="$MY_BASEDIR_VERSION/libexec"
else
ledir="$MY_BASEDIR_VERSION/bin"
fi
elif test -f "$relpkgdata"/english/errmsg.sys -a -x "$MY_PWD/bin/mysqld"
then
MY_BASEDIR_VERSION="$MY_PWD" # Where bin, share and data are
ledir="$MY_PWD/bin" # Where mysqld is
# Check for the directories we would expect from a source install
elif test -f "$relpkgdata"/english/errmsg.sys -a -x "$MY_PWD/libexec/mysqld"
then
MY_BASEDIR_VERSION="$MY_PWD" # Where libexec, share and var are
ledir="$MY_PWD/libexec" # Where mysqld is
# Since we didn't find anything, used the compiled-in defaults
else
MY_BASEDIR_VERSION='@prefix@'
ledir='@libexecdir@'
fi
#
# Second, try to find the data directory
#
# Try where the binary installs put it
if test -d $MY_BASEDIR_VERSION/data/mysql
then
DATADIR=$MY_BASEDIR_VERSION/data
if test -z "$defaults" -a -r "$DATADIR/my.cnf"
then
defaults="--defaults-extra-file=$DATADIR/my.cnf"
fi
# Next try where the source installs put it
elif test -d $MY_BASEDIR_VERSION/var/mysql
then
DATADIR=$MY_BASEDIR_VERSION/var
# Or just give up and use our compiled-in default
else
DATADIR=@localstatedir@
fi
if test -z "$MYSQL_HOME"
then
if test -r "$MY_BASEDIR_VERSION/my.cnf" && test -r "$DATADIR/my.cnf"
then
log_error "WARNING: Found two instances of my.cnf -
$MY_BASEDIR_VERSION/my.cnf and
$DATADIR/my.cnf
IGNORING $DATADIR/my.cnf"
MYSQL_HOME=$MY_BASEDIR_VERSION
elif test -r "$DATADIR/my.cnf"
then
log_error "WARNING: Found $DATADIR/my.cnf
The data directory is a deprecated location for my.cnf, please move it to
$MY_BASEDIR_VERSION/my.cnf"
MYSQL_HOME=$DATADIR
else
MYSQL_HOME=$MY_BASEDIR_VERSION
fi
fi
export MYSQL_HOME
# Get first arguments from the my.cnf file, groups [mysqld] and [mysqld_safe]
# and then merge with the command line arguments
if test -x "$MY_BASEDIR_VERSION/bin/my_print_defaults"
then
print_defaults="$MY_BASEDIR_VERSION/bin/my_print_defaults"
elif test -x ./bin/my_print_defaults
then
print_defaults="./bin/my_print_defaults"
elif test -x @bindir@/my_print_defaults
then
print_defaults="@bindir@/my_print_defaults"
elif test -x @bindir@/mysql_print_defaults
then
print_defaults="@bindir@/mysql_print_defaults"
else
print_defaults="my_print_defaults"
fi
append_arg_to_args () {
args="$args "`shell_quote_string "$1"`
}
args=
SET_USER=2
parse_arguments `$print_defaults $defaults --loose-verbose mysqld server`
if test $SET_USER -eq 2
then
SET_USER=0
fi
parse_arguments `$print_defaults $defaults --loose-verbose mysqld_safe safe_mysqld`
parse_arguments PICK-ARGS-FROM-ARGV "$@"
# A pid file is created for the mysqld_safe process. This file protects the
# server instance resources during race conditions.
safe_pid="$DATADIR/mysqld_safe.pid"
if test -f $safe_pid
then
PID=`cat "$safe_pid"`
if @CHECK_PID@
then
if @FIND_PROC@
then
log_error "A mysqld_safe process already exists"
exit 1
fi
fi
rm -f "$safe_pid"
if test -f "$safe_pid"
then
log_error "Fatal error: Can't remove the mysqld_safe pid file"
exit 1
fi
fi
# Insert pid proerply into the pid file.
ps -e | grep [m]ysqld_safe | awk '{print $1}' | sed -n 1p > $safe_pid
# End of mysqld_safe pid(safe_pid) check.
# Determine what logging facility to use
# Ensure that 'logger' exists, if it's requested
if [ $want_syslog -eq 1 ]
then
my_which logger > /dev/null 2>&1
if [ $? -ne 0 ]
then
log_error "--syslog requested, but no 'logger' program found. Please ensure that 'logger' is in your PATH, or do not specify the --syslog option to mysqld_safe."
rm -f "$safe_pid" # Clean Up of mysqld_safe.pid file.
exit 1
fi
fi
if [ -n "$err_log" -o $want_syslog -eq 0 ]
then
if [ -n "$err_log" ]
then
# mysqld adds ".err" if there is no extension on the --log-error
# argument; must match that here, or mysqld_safe will write to a
# different log file than mysqld
# mysqld does not add ".err" to "--log-error=foo."; it considers a
# trailing "." as an extension
if expr "$err_log" : '.*\.[^/]*$' > /dev/null
then
:
else
err_log="$err_log".err
fi
case "$err_log" in
/* ) ;;
* ) err_log="$DATADIR/$err_log" ;;
esac
else
err_log=$DATADIR/`@HOSTNAME@`.err
fi
append_arg_to_args "--log-error=$err_log"
if [ $want_syslog -eq 1 ]
then
# User explicitly asked for syslog, so warn that it isn't used
log_error "Can't log to error log and syslog at the same time. Remove all --log-error configuration options for --syslog to take effect."
fi
# Log to err_log file
log_notice "Logging to '$err_log'."
logging=file
if [ ! -f "$err_log" ]; then # if error log already exists,
touch "$err_log" # we just append. otherwise,
chmod "$fmode" "$err_log" # fix the permissions here!
fi
else
if [ -n "$syslog_tag" ]
then
# Sanitize the syslog tag
syslog_tag=`echo "$syslog_tag" | sed -e 's/[^a-zA-Z0-9_-]/_/g'`
syslog_tag_mysqld_safe="${syslog_tag_mysqld_safe}-$syslog_tag"
syslog_tag_mysqld="${syslog_tag_mysqld}-$syslog_tag"
fi
log_notice "Logging to syslog."
logging=syslog
fi
USER_OPTION=""
if test -w / -o "$USER" = "root"
then
if test "$user" != "root" -o $SET_USER = 1
then
USER_OPTION="--user=$user"
fi
# Change the err log to the right user, if it is in use
if [ $want_syslog -eq 0 ]; then
touch "$err_log"
chown $user "$err_log"
fi
if test -n "$open_files"
then
ulimit -n $open_files
fi
fi
if test -n "$open_files"
then
append_arg_to_args "--open-files-limit=$open_files"
fi
safe_mysql_unix_port=${mysql_unix_port:-${MYSQL_UNIX_PORT:-@MYSQL_UNIX_ADDR@}}
# Make sure that directory for $safe_mysql_unix_port exists
mysql_unix_port_dir=`dirname $safe_mysql_unix_port`
if [ ! -d $mysql_unix_port_dir ]
then
mkdir $mysql_unix_port_dir
chown $user $mysql_unix_port_dir
chmod 755 $mysql_unix_port_dir
fi
# If the user doesn't specify a binary, we assume name "mysqld"
if test -z "$MYSQLD"
then
MYSQLD=mysqld
fi
if test ! -x "$ledir/$MYSQLD"
then
log_error "The file $ledir/$MYSQLD
does not exist or is not executable. Please cd to the mysql installation
directory and restart this script from there as follows:
./bin/mysqld_safe&
See http://dev.mysql.com/doc/mysql/en/mysqld-safe.html for more information"
rm -f "$safe_pid" # Clean Up of mysqld_safe.pid file.
exit 1
fi
if test -z "$pid_file"
then
pid_file="$DATADIR/`@HOSTNAME@`.pid"
else
case "$pid_file" in
/* ) ;;
* ) pid_file="$DATADIR/$pid_file" ;;
esac
fi
append_arg_to_args "--pid-file=$pid_file"
if test -n "$mysql_unix_port"
then
append_arg_to_args "--socket=$mysql_unix_port"
fi
if test -n "$mysql_tcp_port"
then
append_arg_to_args "--port=$mysql_tcp_port"
fi
if test $niceness -eq 0
then
NOHUP_NICENESS="nohup"
else
NOHUP_NICENESS="nohup nice -$niceness"
fi
# Using nice with no args to get the niceness level is GNU-specific.
# This check could be extended for other operating systems (e.g.,
# BSD could use "nohup sh -c 'ps -o nice -p $$' | tail -1").
# But, it also seems that GNU nohup is the only one which messes
# with the priority, so this is okay.
if nohup nice > /dev/null 2>&1
then
normal_niceness=`nice`
nohup_niceness=`nohup nice 2>/dev/null`
numeric_nice_values=1
for val in $normal_niceness $nohup_niceness
do
case "$val" in
-[0-9] | -[0-9][0-9] | -[0-9][0-9][0-9] | \
[0-9] | [0-9][0-9] | [0-9][0-9][0-9] )
;;
* )
numeric_nice_values=0 ;;
esac
done
if test $numeric_nice_values -eq 1
then
nice_value_diff=`expr $nohup_niceness - $normal_niceness`
if test $? -eq 0 && test $nice_value_diff -gt 0 && \
nice --$nice_value_diff echo testing > /dev/null 2>&1
then
# nohup increases the priority (bad), and we are permitted
# to lower the priority with respect to the value the user
# might have been given
niceness=`expr $niceness - $nice_value_diff`
NOHUP_NICENESS="nice -$niceness nohup"
fi
fi
else
if nohup echo testing > /dev/null 2>&1
then
:
else
# nohup doesn't work on this system
NOHUP_NICENESS=""
fi
fi
# Try to set the core file size (even if we aren't root) because many systems
# don't specify a hard limit on core file size.
if test -n "$core_file_size"
then
ulimit -c $core_file_size
fi
#
# If there exists an old pid file, check if the daemon is already running
# Note: The switches to 'ps' may depend on your operating system
if test -f "$pid_file"
then
PID=`cat "$pid_file"`
if @CHECK_PID@
then
if @FIND_PROC@
then # The pid contains a mysqld process
log_error "A mysqld process already exists"
rm -f "$safe_pid" # Clean Up of mysqld_safe.pid file.
exit 1
fi
fi
rm -f "$pid_file"
if test -f "$pid_file"
then
log_error "Fatal error: Can't remove the pid file:
$pid_file
Please remove it manually and start $0 again;
mysqld daemon not started"
rm -f "$safe_pid" # Clean Up of mysqld_safe.pid file.
exit 1
fi
fi
#
# Uncomment the following lines if you want all tables to be automatically
# checked and repaired during startup. You should add sensible key_buffer
# and sort_buffer values to my.cnf to improve check performance or require
# less disk space.
# Alternatively, you can start mysqld with the "myisam-recover" option. See
# the manual for details.
#
# echo "Checking tables in $DATADIR"
# $MY_BASEDIR_VERSION/bin/myisamchk --silent --force --fast --medium-check $DATADIR/*/*.MYI
# $MY_BASEDIR_VERSION/bin/isamchk --silent --force $DATADIR/*/*.ISM
# Does this work on all systems?
#if type ulimit | grep "shell builtin" > /dev/null
#then
# ulimit -n 256 > /dev/null 2>&1 # Fix for BSD and FreeBSD systems
#fi
cmd="$NOHUP_NICENESS"
for i in "$ledir/$MYSQLD" "$defaults" "--basedir=$MY_BASEDIR_VERSION" \
"--datadir=$DATADIR" "$USER_OPTION"
do
cmd="$cmd "`shell_quote_string "$i"`
done
cmd="$cmd $args"
# Avoid 'nohup: ignoring input' warning
test -n "$NOHUP_NICENESS" && cmd="$cmd < /dev/null"
log_notice "Starting $MYSQLD daemon with databases from $DATADIR"
while true
do
rm -f $safe_mysql_unix_port "$pid_file" # Some extra safety
eval_log_error "$cmd"
if [ $want_syslog -eq 0 -a ! -f "$err_log" ]; then
touch "$err_log" # hypothetical: log was renamed but not
chown $user "$err_log" # flushed yet. we'd recreate it with
chmod "$fmode" "$err_log" # wrong owner next time we log, so set
fi # it up correctly while we can!
if test ! -f "$pid_file" # This is removed if normal shutdown
then
break
fi
if @TARGET_LINUX@ && test $KILL_MYSQLD -eq 1
then
# Test if one process was hanging.
# This is only a fix for Linux (running as base 3 mysqld processes)
# but should work for the rest of the servers.
# The only thing is ps x => redhat 5 gives warnings when using ps -x.
# kill -9 is used or the process won't react on the kill.
numofproces=`ps xaww | grep -v "grep" | grep "$ledir/$MYSQLD\>" | grep -c "pid-file=$pid_file"`
log_notice "Number of processes running now: $numofproces"
I=1
while test "$I" -le "$numofproces"
do
PROC=`ps xaww | grep "$ledir/$MYSQLD\>" | grep -v "grep" | grep "pid-file=$pid_file" | sed -n '$p'`
for T in $PROC
do
break
done
# echo "TEST $I - $T **"
if kill -9 $T
then
log_error "$MYSQLD process hanging, pid $T - killed"
else
break
fi
I=`expr $I + 1`
done
fi
log_notice "mysqld restarted"
done
log_notice "mysqld from pid file $pid_file ended"
rm -f "$safe_pid" # Some Extra Safety. File is deleted
# once the mysqld process ends.