oom_of_squid
用於監控服務器上的squid進程，保證對用戶服務的穩定。
此進程意欲以守護進程模式運行，可以這樣使用：./oom_of_squid &
主要功能：
監控配置過的每個squid進程，保證它的內存佔用看似正常；
確保進程PID和PID文件一致；
squid進程掛掉後啓動之；
squid進程佔用的內存超過設置的閥值則重啓之；
系統可用內存低於閥值，則選擇一個squid進程重啓之；
默認每3秒檢查一次；
注意事項：
`echo squid7[0-9].conf`在gentoo和CentOS上的行爲不一樣，到CentOS下應用可能需要改爲`ls squid7[0-9].conf`；
/proc/$PID/comm 在較新的內核上纔有這個接口，使用舊內核需要另想辦法；
/etc/init.d/squid_multi_instance是我重寫過的進程管理腳本（還有bug），使用“標準”腳本的人需要自己改一下；
系統可用內存閥值應該適當高於系統自身的oom閥值；
用於全內存式緩存更好，因爲不用顧忌狀態文件的同步。
我設置的配置文件中，squid7[0-9].conf是carp-child，用於緩存文件；squid8[0-9].conf是carp-parent，用於決定URL在集羣中的分佈。
附件爲腳本。望用的上的同仁多提寶貴意見！！！
|
#!/bin/bash
# 功能:
# 在系統可用內存很小時選取一個squid進程重啓
# 檢查各進程健康狀況
MAIL="[email protected]" # 設置自己的郵箱
MAIL_FROM="[email protected]"
CHECK_INTERVAL="3s" # 檢查時間間隔
THRESHOLD_LOW_MEM="90" # MB,低內存閥值; 必須大於vm.min_free_kbytes
THRESHOLD_CHILD_MAX_MEM="200" # MB,child持有內存大於此值就重啓
THRESHOLD_EMERG="100" # MB,LEVEL="emerg"時重啓進程的閥值
LEVEL="normal" # 內存級別,可用內存較低時設置爲emerg; 正常情況下應該是空變量
LOG="$0.log"
# 探測配置文件,70-79之間的端口爲parent,80-87爲child
CONFIG_LOCATION="/etc/squid/"
PARENT_CONF=$(cd $CONFIG_LOCATION; echo squid7[0-9].conf)
CHILD_CONF=$(cd $CONFIG_LOCATION; echo squid8[0-9].conf)
ALL_CONF="$PARENT_CONF $CHILD_CONF"
# disable killing by oom
echo "-17" > /proc/self/oom_adj
all_free_mem() {
free -m | awk '/^Mem:/ {print ($4 + $6 + $7)}' # 單位是MB
}
squid_instance_mem() {
INSTANCE=${CONF%.conf}
PID_FILE="/var/run/$INSTANCE.pid"
[[ -e $PID_FILE ]] && PID_FROM_FILE=`cat $PID_FILE` || PID_FROM_FILE=""
# 檢查名稱和進程號是否對應
if [ -e $PID_FILE ]; then
[[ x$PID_FROM_FILE != "x" ]] && grep -q squid /proc/$PID_FROM_FILE/comm || false
else
echo "$INSTANCE pid file not match to squid" >> $LOG
fi
# 佔用的內存數量
RSS=`awk '/^VmRSS:/ {printf "%d\n", $2/1024}' /proc/$PID_FROM_FILE/status`
}
keep_processes_health() {
# 檢查所有進程,如果掛了,修正一下
for CONF in $ALL_CONF; do
INSTANCE=${CONF%.conf}
PID_FILE="/var/run/$INSTANCE.pid"
STATE_FILE="/var/lib/init.d/started/$INSTANCE"
[[ -e $PID_FILE ]] && PID_FROM_FILE=`cat $PID_FILE` || PID_FROM_FILE=""
PID_RUNNING=`ps axo user,pid,cmd | awk '/^squid/ && /'$CONF'/ {print $2}'`
if [ x$PID_RUNNING = "x" ]; then
# 進程不存在: 啓動進程
process_state="not_running"
echo "`date +%F\ %T` $INSTANCE state is $process_state, restarted" >> $LOG
[[ -e $PID_FILE ]] && /bin/rm -f $PID_FILE
[[ -e $STATE_FILE ]] && /bin/rm -f $STATE_FILE
/etc/init.d/squid_multi_instance start ${INSTANCE#squid} >/dev/null 2>&1
notify_admin &
elif [ x$PID_FROM_FILE = "x" ]; then
# pid文件有問題: 修正
process_state="bad_pid_file"
if echo $PID_RUNNING > $PID_FILE; then
process_state="good"
echo "fixed pid file of $INSTANCE at `date +%F\ %T`" >> $LOG
else
echo "can not write $PID_FILE" >> $LOG
process_state="pid_file_not_writeable"
notify_admin &
fi
elif [ x$PID_FROM_FILE != x$PID_RUNNING ]; then
# 什麼情況下會這樣呢...
process_state="pid_not_equal"
if echo $PID_RUNNING > $PID_FILE; then
echo "fixed $process_state of $INSTANCE at `date +%F\ %T`" >> $LOG
notify_admin &
else
echo "can not write $PID_FILE" >> $LOG
process_state="$process_state pid_file_not_writeable"
notify_admin &
fi
elif [ x$PID_FROM_FILE = x$PID_RUNNING ]; then
# 運行良好
process_state="good"
continue
else
process_state="unknow"
notify_admin &
fi
done
unset process_state
}
restart_process() {
INSTANCE=${CONF%.conf}
/etc/init.d/squid_multi_instance restart ${INSTANCE#squid} >/dev/null 2>&1
notify_admin &
}
pick_and_restart_parent() {
# 重啓佔用內存最大的進程
for CONF in $PARENT_CONF; do
squid_instance_mem
PARENT_MEM="$PARENT_MEM\n$RSS $CONF"
done
PARENT_MEM_MAX=`echo -e $PARENT_MEM | sort -n | tail -1`
if [ -n $PARENT_MEM_MAX ]; then
CONF=`echo $PARENT_MEM_MAX | awk '{print $2}'`
restart_process
echo "restarted parent ${CONF%.conf} at `date +%F\ %T`" >> $LOG
else
echo "unknow error in pick_and_restart_parent" >> $LOG
fi
}
pick_and_restart_child() {
# 正常情況下重啓內存大於300M的進程; 緊急情況下重啓大於100M的進程
for CONF in $CHILD_CONF; do
squid_instance_mem
if [ $RSS -gt $THRESHOLD_CHILD_MAX_MEM ]; then
process_state="over_THRESHOLD_CHILD_MAX_MEM"
echo "$process_state restarted child ${CONF%.conf} at `date +%F\ %T`" >> $LOG
restart_process
else
true
fi
done
if [ x$LEVEL = "xemerg" ]; then
for CONF in $CHILD_CONF; do
squid_instance_mem
[[ $RSS -gt $THRESHOLD_EMERG ]] && (process_state="over_THRESHOLD_EMERG"; restart_process)
done
else
true
fi
}
pick_and_restart_one() {
# restart a child first
pick_and_restart_child
# memory still low, restart a parent
[[ `all_free_mem` -lt $THRESHOLD_LOW_MEM ]] && pick_and_restart_parent
# memory still low, restart all child
[[ `all_free_mem` -lt $THRESHOLD_LOW_MEM ]] && \
(CONF="child"; restart_process)
}
notify_admin() {
sendmail -t -f $MAIL_FROM <<EOF
To: $MAIL
From: $MAIL_FROM
Subject: $INSTANCE on `hostname` restarted
重啓過了 `hostname` 上的 $INSTANCE
進程 $INSTANCE 的最後狀態爲: $process_state
當前系統總剩餘內存爲: `all_free_mem`
.
EOF
}
while true; do
keep_processes_health
# child佔用內存大於THRESHOLD_CHILD_MAX_MEM就重啓,不等總內存過低
pick_and_restart_child
if [ `all_free_mem` -lt $THRESHOLD_LOW_MEM ]; then
LEVEL="emerg"
echo "low memory at `date +%F\ %T`" >> $LOG
pick_and_restart_one
else
LEVEL="normal"
sleep $CHECK_INTERVAL
fi
done
# vim: set sw=4 ts=4:
|
