核心服務器上跑了一堆的腳本、程序，難免有時候會出現(xiàn)僵尸進程，死不死活不活的在那里占用資源，最初只是寫了個根據(jù)關鍵字查殺進程的linux shell腳本，后來發(fā)現(xiàn)很多時候進程死在那里的時候其實是內(nèi)部調(diào)用子進程的時候出現(xiàn)了問題，這時候光殺父進程根本沒解決根本問題。比如說rsync的時候通過ssh來連接，rsync本身沒問題，但可能ssh死掉了。因此重新寫了腳本，遞歸查找子進程。
    代碼如下:
    #!/bin/sh
    # 遞歸找到導致進程僵死的最底層子進程并殺除.
    ParentProcessID=$1;
    if [ "x${ParentProcessID}" = "x" ] ; then
    echo "Please Supply the top Parent Process ID to be killed!"
    echo "Usage:sh $0 PID [-v]"
    echo "PID The Parent Process ID as root"
    echo "-v is this argument supplied,no real kill operation will be performed,only process tree be show."
    exit 1
    fi
    let IsRealKillDo=1;
    if [ "x$2" = "x-v" ] ; then
    let IsRealKillDo=0;
    fi
    echo "Begin Kill the Leaf Process of process ${ParentProcessID}" >&2
    killpidList=""
    function loopNextSubProcess(){
    local nParentProcessID=$1
    local tmpPidList=""
    tmpPidList=`ps -A --format='%p%PisParent' --width 2048 -w --sort pid|grep "${nParentProcessID}isParent"|grep -v grep|grep -v "$$" | awk '{ printf $1 }'`
    ps --format='%p%P%a' --width 2048 -w -p ${nParentProcessID}|grep -v grep|grep -v "$$" >&2
    if [ "x${tmpPidList}" = "x" ] ; then
    echo "****Got One Leaf = [${nParentProcessID}]****" >&2
    killpidList="${killpidList}\n${nParentProcessID}"
    return
    fi
    for theNextPid in ${tmpPidList} ; do
    loopNextSubProcess ${theNextPid}
    done
    }
    loopNextSubProcess ${ParentProcessID}
    if [ ${IsRealKillDo} -eq 1 -a "x${killpidList}" != "x" ] ; then
    for curpid in `echo -e ${killpidList}` ; do
    if [ "x${curpid}" != "x" ] ; then
    echo "kill -9 ${curpid}"
    kill -9 ${curpid}
    fi
    done
    else
    echo -e ${killpidList}
    fi