mirror of
https://github.com/mailcow/mailcow-dockerized.git
synced 2026-02-14 16:30:38 +00:00
Merge branch 'staging' into feat/valkey
This commit is contained in:
@@ -1,4 +1,4 @@
|
||||
FROM alpine:3.20
|
||||
FROM alpine:3.21
|
||||
|
||||
LABEL maintainer = "The Infrastructure Company GmbH <info@servercow.de>"
|
||||
|
||||
@@ -16,7 +16,6 @@ RUN apk add --update \
|
||||
fcgi \
|
||||
openssl \
|
||||
nagios-plugins-mysql \
|
||||
nagios-plugins-dns \
|
||||
nagios-plugins-disk \
|
||||
bind-tools \
|
||||
redis \
|
||||
@@ -32,9 +31,11 @@ RUN apk add --update \
|
||||
tzdata \
|
||||
whois \
|
||||
&& curl https://raw.githubusercontent.com/mludvig/smtp-cli/v3.10/smtp-cli -o /smtp-cli \
|
||||
&& chmod +x smtp-cli
|
||||
&& chmod +x smtp-cli \
|
||||
&& mkdir /usr/lib/mailcow
|
||||
|
||||
COPY watchdog.sh /watchdog.sh
|
||||
COPY check_mysql_slavestatus.sh /usr/lib/nagios/plugins/check_mysql_slavestatus.sh
|
||||
COPY check_dns.sh /usr/lib/mailcow/check_dns.sh
|
||||
|
||||
CMD ["/watchdog.sh"]
|
||||
|
||||
39
data/Dockerfiles/watchdog/check_dns.sh
Executable file
39
data/Dockerfiles/watchdog/check_dns.sh
Executable file
@@ -0,0 +1,39 @@
|
||||
#!/bin/sh
|
||||
|
||||
while getopts "H:s:" opt; do
|
||||
case "$opt" in
|
||||
H) HOST="$OPTARG" ;;
|
||||
s) SERVER="$OPTARG" ;;
|
||||
*) echo "Usage: $0 -H host -s server"; exit 3 ;;
|
||||
esac
|
||||
done
|
||||
|
||||
if [ -z "$SERVER" ]; then
|
||||
echo "No DNS Server provided"
|
||||
exit 3
|
||||
fi
|
||||
|
||||
if [ -z "$HOST" ]; then
|
||||
echo "No host to test provided"
|
||||
exit 3
|
||||
fi
|
||||
|
||||
# run dig and measure the time it takes to run
|
||||
START_TIME=$(date +%s%3N)
|
||||
dig_output=$(dig +short +timeout=2 +tries=1 "$HOST" @"$SERVER" 2>/dev/null)
|
||||
dig_rc=$?
|
||||
dig_output_ips=$(echo "$dig_output" | grep -E '^[0-9.]+$' | sort | paste -sd ',' -)
|
||||
END_TIME=$(date +%s%3N)
|
||||
ELAPSED_TIME=$((END_TIME - START_TIME))
|
||||
|
||||
# validate and perform nagios like output and exit codes
|
||||
if [ $dig_rc -ne 0 ] || [ -z "$dig_output" ]; then
|
||||
echo "Domain $HOST was not found by the server"
|
||||
exit 2
|
||||
elif [ $dig_rc -eq 0 ]; then
|
||||
echo "DNS OK: $ELAPSED_TIME ms response time. $HOST returns $dig_output_ips"
|
||||
exit 0
|
||||
else
|
||||
echo "Unknown error"
|
||||
exit 3
|
||||
fi
|
||||
@@ -49,7 +49,7 @@
|
||||
# 2013101601 Optical clean up #
|
||||
# 2013101602 Rewrite help output #
|
||||
# 2013101700 Handle Slave IO in 'Connecting' state #
|
||||
# 2013101701 Minor changes in output, handling UNKWNON situations now #
|
||||
# 2013101701 Minor changes in output, handling UNKNOWN situations now #
|
||||
# 2013101702 Exit CRITICAL when Slave IO in Connecting state #
|
||||
# 2013123000 Slave_SQL_Running also matched Slave_SQL_Running_State #
|
||||
# 2015011600 Added 'moving' check to catch possible connection issues #
|
||||
@@ -131,10 +131,10 @@ elif [[ -n "${socket}" && (-z "${user}" || -z "${password}") ]]; then
|
||||
fi
|
||||
|
||||
# Connect to the DB server and store output in vars
|
||||
if [[ -n $socket ]]; then
|
||||
ConnectionResult=$(mysql ${optfile} ${socket} ${user} -e "show slave ${connection} status\G" 2>&1)
|
||||
if [[ -n $socket ]]; then
|
||||
ConnectionResult=$(mariadb --skip-ssl ${optfile} ${socket} ${user} -e "show slave ${connection} status\G" 2>&1)
|
||||
else
|
||||
ConnectionResult=$(mysql ${optfile} ${host} ${port} ${user} -e "show slave ${connection} status\G" 2>&1)
|
||||
ConnectionResult=$(mariadb --skip-ssl ${optfile} ${host} ${port} ${user} -e "show slave ${connection} status\G" 2>&1)
|
||||
fi
|
||||
|
||||
if [ -z "`echo "${ConnectionResult}" |grep Slave_IO_State`" ]; then
|
||||
@@ -178,33 +178,33 @@ if [ ${check} = ${ok} ] && [ ${checkio} = ${ok} ]; then
|
||||
then echo "CRITICAL: Slave is ${delayinfo} seconds behind Master | delay=${delayinfo}s"; exit ${STATE_CRITICAL}
|
||||
elif [[ ${delayinfo} -ge ${warn_delay} ]]
|
||||
then echo "WARNING: Slave is ${delayinfo} seconds behind Master | delay=${delayinfo}s"; exit ${STATE_WARNING}
|
||||
else
|
||||
else
|
||||
# Everything looks OK here but now let us check if the replication is moving
|
||||
if [[ -n ${moving} ]] && [[ -n ${tmpfile} ]] && [[ $readpos -eq $execpos ]]
|
||||
then
|
||||
#echo "Debug: Read pos is $readpos - Exec pos is $execpos"
|
||||
then
|
||||
#echo "Debug: Read pos is $readpos - Exec pos is $execpos"
|
||||
# Check if tmp file exists
|
||||
curtime=`date +%s`
|
||||
if [[ -w $tmpfile ]]
|
||||
then
|
||||
if [[ -w $tmpfile ]]
|
||||
then
|
||||
tmpfiletime=`date +%s -r $tmpfile`
|
||||
if [[ `expr $curtime - $tmpfiletime` -gt ${moving} ]]
|
||||
then
|
||||
exectmp=`cat $tmpfile`
|
||||
#echo "Debug: Exec pos in tmpfile is $exectmp"
|
||||
if [[ $exectmp -eq $execpos ]]
|
||||
then
|
||||
then
|
||||
# The value read from the tmp file and from db are the same. Replication hasnt moved!
|
||||
echo "WARNING: Slave replication has not moved in ${moving} seconds. Manual check required."; exit ${STATE_WARNING}
|
||||
else
|
||||
else
|
||||
# Replication has moved since the tmp file was written. Delete tmp file and output OK.
|
||||
rm $tmpfile
|
||||
echo "OK: Slave SQL running: ${check} Slave IO running: ${checkio} / master: ${masterinfo} / slave is ${delayinfo} seconds behind master | delay=${delayinfo}s"; exit ${STATE_OK};
|
||||
fi
|
||||
else
|
||||
else
|
||||
echo "OK: Slave SQL running: ${check} Slave IO running: ${checkio} / master: ${masterinfo} / slave is ${delayinfo} seconds behind master | delay=${delayinfo}s"; exit ${STATE_OK};
|
||||
fi
|
||||
else
|
||||
else
|
||||
echo "$execpos" > $tmpfile
|
||||
echo "OK: Slave SQL running: ${check} Slave IO running: ${checkio} / master: ${masterinfo} / slave is ${delayinfo} seconds behind master | delay=${delayinfo}s"; exit ${STATE_OK};
|
||||
fi
|
||||
|
||||
@@ -1,5 +1,10 @@
|
||||
#!/bin/bash
|
||||
|
||||
if [ "${DEV_MODE}" != "n" ]; then
|
||||
echo -e "\e[31mEnabled Debug Mode\e[0m"
|
||||
set -x
|
||||
fi
|
||||
|
||||
trap "exit" INT TERM
|
||||
trap "kill 0" EXIT
|
||||
|
||||
@@ -234,7 +239,7 @@ external_checks() {
|
||||
diff_c=0
|
||||
THRESHOLD=${EXTERNAL_CHECKS_THRESHOLD}
|
||||
# Reduce error count by 2 after restarting an unhealthy container
|
||||
GUID=$(mysql -u${DBUSER} -p${DBPASS} ${DBNAME} -e "SELECT version FROM versions WHERE application = 'GUID'" -BN)
|
||||
GUID=$(mariadb --skip-ssl -u${DBUSER} -p${DBPASS} ${DBNAME} -e "SELECT version FROM versions WHERE application = 'GUID'" -BN)
|
||||
trap "[ ${err_count} -gt 1 ] && err_count=$(( ${err_count} - 2 ))" USR1
|
||||
while [ ${err_count} -lt ${THRESHOLD} ]; do
|
||||
err_c_cur=${err_count}
|
||||
@@ -297,7 +302,7 @@ unbound_checks() {
|
||||
touch /tmp/unbound-mailcow; echo "$(tail -50 /tmp/unbound-mailcow)" > /tmp/unbound-mailcow
|
||||
host_ip=$(get_container_ip unbound-mailcow)
|
||||
err_c_cur=${err_count}
|
||||
/usr/lib/nagios/plugins/check_dns -s ${host_ip} -H stackoverflow.com 2>> /tmp/unbound-mailcow 1>&2; err_count=$(( ${err_count} + $? ))
|
||||
/usr/lib/mailcow/check_dns.sh -s ${host_ip} -H stackoverflow.com 2>> /tmp/unbound-mailcow 1>&2; err_count=$(( ${err_count} + $? ))
|
||||
DNSSEC=$(dig com +dnssec | egrep 'flags:.+ad')
|
||||
if [[ -z ${DNSSEC} ]]; then
|
||||
echo "DNSSEC failure" 2>> /tmp/unbound-mailcow 1>&2
|
||||
@@ -445,6 +450,31 @@ postfix_checks() {
|
||||
return 1
|
||||
}
|
||||
|
||||
postfix-tlspol_checks() {
|
||||
err_count=0
|
||||
diff_c=0
|
||||
THRESHOLD=${POSTFIX_TLSPOL_THRESHOLD}
|
||||
# Reduce error count by 2 after restarting an unhealthy container
|
||||
trap "[ ${err_count} -gt 1 ] && err_count=$(( ${err_count} - 2 ))" USR1
|
||||
while [ ${err_count} -lt ${THRESHOLD} ]; do
|
||||
touch /tmp/postfix-tlspol-mailcow; echo "$(tail -50 /tmp/postfix-tlspol-mailcow)" > /tmp/postfix-tlspol-mailcow
|
||||
host_ip=$(get_container_ip postfix-tlspol-mailcow)
|
||||
err_c_cur=${err_count}
|
||||
/usr/lib/nagios/plugins/check_tcp -4 -H ${host_ip} -p 8642 2>> /tmp/postfix-tlspol-mailcow 1>&2; err_count=$(( ${err_count} + $? ))
|
||||
[ ${err_c_cur} -eq ${err_count} ] && [ ! $((${err_count} - 1)) -lt 0 ] && err_count=$((${err_count} - 1)) diff_c=1
|
||||
[ ${err_c_cur} -ne ${err_count} ] && diff_c=$(( ${err_c_cur} - ${err_count} ))
|
||||
progress "Postfix TLS Policy companion" ${THRESHOLD} $(( ${THRESHOLD} - ${err_count} )) ${diff_c}
|
||||
if [[ $? == 10 ]]; then
|
||||
diff_c=0
|
||||
sleep 1
|
||||
else
|
||||
diff_c=0
|
||||
sleep $(( ( RANDOM % 60 ) + 20 ))
|
||||
fi
|
||||
done
|
||||
return 1
|
||||
}
|
||||
|
||||
clamd_checks() {
|
||||
err_count=0
|
||||
diff_c=0
|
||||
@@ -922,6 +952,18 @@ PID=$!
|
||||
echo "Spawned mailq_checks with PID ${PID}"
|
||||
BACKGROUND_TASKS+=(${PID})
|
||||
|
||||
(
|
||||
while true; do
|
||||
if ! postfix-tlspol_checks; then
|
||||
log_msg "Postfix TLS Policy hit error limit"
|
||||
echo postfix-tlspol-mailcow > /tmp/com_pipe
|
||||
fi
|
||||
done
|
||||
) &
|
||||
PID=$!
|
||||
echo "Spawned postfix-tlspol_checks with PID ${PID}"
|
||||
BACKGROUND_TASKS+=(${PID})
|
||||
|
||||
(
|
||||
while true; do
|
||||
if ! dovecot_checks; then
|
||||
@@ -994,6 +1036,7 @@ PID=$!
|
||||
echo "Spawned cert_checks with PID ${PID}"
|
||||
BACKGROUND_TASKS+=(${PID})
|
||||
|
||||
if [[ "${SKIP_OLEFY}" =~ ^([nN][oO]|[nN])+$ ]]; then
|
||||
(
|
||||
while true; do
|
||||
if ! olefy_checks; then
|
||||
@@ -1005,6 +1048,7 @@ done
|
||||
PID=$!
|
||||
echo "Spawned olefy_checks with PID ${PID}"
|
||||
BACKGROUND_TASKS+=(${PID})
|
||||
fi
|
||||
|
||||
(
|
||||
while true; do
|
||||
|
||||
Reference in New Issue
Block a user