服务器 
首页 > 服务器 > 浏览文章

cloudstack下libvirtd服务无响应问题

(编辑:jimmy 日期: 2024/11/26 浏览:3 次 )

在cloudstack4.5.2版本下,偶尔出现libvirtd服务无响应的情况,导致virsh命令无法使用,同时伴随cloudstack master丢失该slave主机连接的情况。最初怀疑是libvirtd服务或版本的问题,经过分析和排查最终确定是cloudstack-agent的问题。但是在官网上并没有找到类似的bug提交,该问题可能还存在于更高的版本,需要时间进一步从根本上分析。下面是该问题的处理过程,在此记录下,关注和使用cloudstack的朋友可以参考。

众所周知,cloudstack的社区热度远不如openstack,为什么还要选择clcoudstack?这个问题以后有机会再和大家聊。言归正传。

环境交代

宿主机操作系统:centos6.5x64(2.6.32-431.el6.x86_64)
cloudstack版本:4.5.2
libvirt版本:libvirt-0.10.2-54.el6_7.2.x86_64

问题描述

通过cloudstackapi listHosts报警信息显示:
node5.cloud.rtmap:192.168.14.20 state is Down at 2016-05-13T07:19:04+0800
#有关cloudstackapi的使用方法在其它文章中总结,不在此处说明。

登陆问题宿主服务器检查:

[root@node5 log]#virsh list --all

没有响应ctrl^c退出

这时的vm可以正常工作,但处于失控状态

尝试重启启动libvirtd服务:

[root@node5 log]# service libvirtd stop

正在关闭 libvirtd 守护进程:                               [失败]  #无法关闭libvirtd服务

尝试重启启动cloudstack-agent服务:

[root@node5 libvirt]# service cloudstack-agent restart
Stopping Cloud Agent: 
Starting Cloud Agent: 

libvirtd故障依旧

简单维护

[root@node5 ping]# libvirtd -d -l --config /etc/libvirt/libvirtd.conf 

libvirtd:错误:Unable to initialize network sockets。查看 /var/log/messages 或者运行不带 --daemon 的命令查看更多信息。

[root@node5 log]# libvirtd -d

可以执行成功,这时执行virsh list --all 可以查看和操作vm

[root@node5 log]#virsh list --all
Id  名称             状态
----------------------------------------------------
 2   i-4-185-VM           running

虽然vm运行正常,现在也可以通过命令正常管理了。但是对于cloudstack平台而言,宿主机处于down状态,vm处于失控状态。

临时解决办法是在其它大的升级和维护过程中重启服务器解决,根本解决还要具体问题具体分析。

分析与排查

检查进程

[root@node5 log]# ps ax |grep libvirtd
 6485 "htmlcode">
[root@node5 log]# top -p 6485
top -p 6485
top - 09:19:41 up 12 days, 22:27, 1 user, load average: 3.05, 5.07, 6.64
Tasks:  1 total,  0 running,  1 sleeping,  0 stopped,  0 zombie
Cpu(s): 4.8%us, 1.4%sy, 0.0%ni, 93.1%id, 0.6%wa, 0.0%hi, 0.1%si, 0.0%st
Mem: 264420148k total, 182040780k used, 82379368k free,  834232k buffers
Swap: 8388600k total,    92k used, 8388508k free, 100453708k cached

  PID USER   PR NI VIRT RES SHR S %CPU %MEM  TIME+ COMMAND
 6485 root   20  0 984m 12m 4440 R 100.2 0.0 844:22.68 libvirtd       #cpu占用100%,无法释放,影响系统稳定性

杀进程

[root@node5 log]# kill -9 6485
[root@node5 log]# kill -9 6485
[root@master log]# ps ax |grep libvirtd  #检查进程依然存在
 6485 "htmlcode">
[root@node5 log]#strace -f libvirtd
[pid 107570] close(23058)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23059)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23060)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23061)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23062)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23063)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23064)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23065)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23066)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23067)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23068)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23069)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23070)        = -1 EBADF (Bad file descriptor)
[pid 107570] close(23071)        = -1 EBADF (Bad file descriptor)
^C[pid 107570] close(23072 <unfinished ...>
Process 107559 detached
Process 107560 detached
Process 107561 detached
Process 107562 detached
Process 107563 detached
Process 107564 detached
Process 107565 detached
Process 107566 detached
Process 107567 detached
Process 107568 detached
Process 107569 detached
Process 107570 detached

父进程6485在不断的产生和关闭子进程,并返回错误信息。Bad file descriptor的原因(如何触发的,谁触发的)? 循环为何无法退出?问题如何再现?

获得更多的线索

官方文档(libvirtd各种故障诊断记录和解决办法非常详尽)
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Virtualization_Deployment_and_Administration_Guide/sect-Troubleshooting-Common_libvirt_errors_and_troubleshooting.html#sect-libvirtd_failed_to_start

开启系统日志

Change libvirt's logging in /etc/libvirt/libvirtd.conf by enabling the line below. To enable the setting the line, open the /etc/libvirt/libvirtd.conf file in a text editor, remove the hash (or #) symbol from the beginning of the following line, and save the change:
log_outputs="3:syslog:libvirtd"

参照配置,重启服务器等待下次故障观察日志

......

Jun 1 12:42:26 node5 abrtd: New client connected
Jun 1 12:42:26 node5 abrtd: Directory 'pyhook-2016-06-01-12:42:26-70065' creation detected
Jun 1 12:42:26 node5 abrt-server[70066]: Saved Python crash dump of pid 70065 to /var/spool/abrt/pyhook-2016-06-01-12:42:26-70065
Jun 1 12:42:26 node5 abrtd: Package 'cloudstack-common' isn't signed with proper key
Jun 1 12:42:26 node5 abrtd: 'post-create' on '/var/spool/abrt/pyhook-2016-06-01-12:42:26-70065' exited with 1
Jun 1 12:42:26 node5 abrtd: Deleting problem directory '/var/spool/abrt/pyhook-2016-06-01-12:42:26-70065'
Jun 1 12:43:26 node5 abrt: detected unhandled Python exception in '/usr/share/cloudstack-common/scripts/vm/network/security_group.py'
......

Jun 6 10:36:21 node5 libvirtd: 102840: warning : qemuDomainObjBeginJobInternal:878 : Cannot start job (modify, none) for domain i-4-30-VM; current job is (modify, none) owned by (102925, 0)
Jun 6 10:36:21 node5 libvirtd: 102840: error : qemuDomainObjBeginJobInternal:883 : Timed out during operation: cannot acquire state change lock
Jun 6 10:39:59 node5 libvirtd: 114071: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 6 10:39:59 node5 libvirtd: 114071: error : virNetSocketNewListenTCP:312 : Unable to bind to port: 地址已在使用
Jun 6 10:40:46 node5 libvirtd: 114147: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 6 10:40:46 node5 libvirtd: 114147: error : virNetSocketNewListenTCP:312 : Unable to bind to port: 地址已在使用
Jun 6 10:42:15 node5 libvirtd: 114204: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 6 10:42:15 node5 libvirtd: 114204: error : virNetSocketNewListenTCP:312 : Unable to bind to port: 地址已在使用
Jun 6 10:47:05 node5 libvirtd: 114375: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 6 10:47:05 node5 libvirtd: 114375: error : virNetSocketNewListenTCP:312 : Unable to bind to port: 地址已在使用
Jun 6 10:47:23 node5 libvirtd: 114412: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 6 10:47:23 node5 libvirtd: 114412: error : virNetSocketNewListenTCP:312 : Unable to bind to port: 地址已在使用
......

Jun 12 03:08:02 node5 rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" x-pid="3111" x-info="http://www.rsyslog.com"] rsyslogd was HUPed
Jun 12 09:20:40 node5 libvirtd: 72575: info : libvirt version: 0.10.2, package: 54.el6_7.2 (CentOS BuildSystem <http://bugs.centos.org>, 2015-11-10-10:25:08, c6b9.bsys.dev.centos.org)
Jun 12 09:20:40 node5 libvirtd: 72575: error : virPidFileAcquirePath:410 : Failed to acquire pid file '/var/run/libvirtd.pid': 资源暂时不可用

并未获得致命错误和更多线索。(该日志配置选项还是很有必要打开的,很多问题都可以通过它来定位)

解决过程

解决思路

尝试和找到终止进程、重启服务的方法
提交bug,等待补丁升级
分析源代码,再现问题,解决问题(投入研发和时间)
由于不能再现问题,还是从简入繁吧。触发这些子进程的元凶是谁?还是cloudstack-agent的嫌疑最大,但之前重启过该服务并没有解决问题,那么agent服务是怎么一回事呢?

看下启动脚本可以基本了解,

[root@node5 libvirt]# cat /etc/rc.d/init.d/cloudstack-agent

#!/bin/bash

# chkconfig: 35 99 10
# description: Cloud Agent

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
#  http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.

# WARNING: if this script is changed, then all other initscripts MUST BE changed to match it as well

. /etc/rc.d/init.d/functions

# set environment variables

SHORTNAME=$(basename $0 | sed -e 's/^[SK][0-9][0-9]//')
PIDFILE=/var/run/"$SHORTNAME".pid
LOCKFILE=/var/lock/subsys/"$SHORTNAME"
LOGDIR=/var/log/cloudstack/agent
LOGFILE=${LOGDIR}/agent.log
PROGNAME="Cloud Agent"
CLASS="com.cloud.agent.AgentShell"
JSVC=`which jsvc 2>/dev/null`;

# exit if we don't find jsvc
if [ -z "$JSVC" ]; then
  echo no jsvc found in path;
  exit 1;
fi

unset OPTIONS
[ -r /etc/sysconfig/"$SHORTNAME" ] && source /etc/sysconfig/"$SHORTNAME"

# The first existing directory is used for JAVA_HOME (if JAVA_HOME is not defined in $DEFAULT)
JDK_DIRS="/usr/lib/jvm/jre /usr/lib/jvm/java-7-openjdk /usr/lib/jvm/java-7-openjdk-i386 /usr/lib/jvm/java-7-openjdk-amd64 /usr/lib/jvm/java-6-openjdk /usr/lib/jvm/java-6-openjdk-i386 /usr/lib/jvm/java-6-openjdk-amd64 /usr/lib/jvm/java-6-sun"

for jdir in $JDK_DIRS; do
  if [ -r "$jdir/bin/java" -a -z "${JAVA_HOME}" ]; then
    JAVA_HOME="$jdir"
  fi
done
export JAVA_HOME

ACP=`ls /usr/share/cloudstack-agent/lib/*.jar | tr '\n' ':' | sed s'/.$//'`
PCP=`ls /usr/share/cloudstack-agent/plugins/*.jar 2>/dev/null | tr '\n' ':' | sed s'/.$//'`

# We need to append the JSVC daemon JAR to the classpath
# AgentShell implements the JSVC daemon methods
export CLASSPATH="/usr/share/java/commons-daemon.jar:$ACP:$PCP:/etc/cloudstack/agent:/usr/share/cloudstack-common/scripts"

start() {
  echo -n $"Starting $PROGNAME: "
  if hostname --fqdn >/dev/null 2>&1 ; then
    $JSVC -Xms256m -Xmx2048m -cp "$CLASSPATH" -pidfile "$PIDFILE"       -errfile $LOGDIR/cloudstack-agent.err -outfile $LOGDIR/cloudstack-agent.out $CLASS
    RETVAL=$"$PROGNAME". > /dev/stderr
    RETVAL=9
  fi
  [ $RETVAL = 0 ] && touch ${LOCKFILE}
  return $RETVAL
}

stop() {
  echo -n $"Stopping $PROGNAME: "
  $JSVC -pidfile "$PIDFILE" -stop $CLASS
  RETVAL=$"$1" in
  start)
    start
    ;;
  stop)
    stop
    ;;
  status)
    status -p ${PIDFILE} $SHORTNAME
    RETVAL=$"Usage: $SHORTNAME {start|stop|restart|condrestart|status|help}"
  RETVAL=3
esac

exit $RETVAL

[root@node5 libvirt]# ps ax |grep jsvc.exec
 6655 "htmlcode">
[root@node5 bin]# service cloudstack-agent status
cloudstack-agent (pid 6657) 正在运行...
[root@node5 bin]# service cloudstack-agent stop
Stopping Cloud Agent:

[root@node5 bin]# service cloudstack-agent status
cloudstack-agent (pid 6657) 正在运行..

ps ax |grep jsvc.exec 也验证了进程依然存在

眼前一亮的同时,也发现了之前使用restart带来的问题,stop不成功的问题被掩盖了~~~有没有懊恼? 不过来不及反思,接下来的问题还远不是这么简单......

[root@node5 bin]# kill -9 6655 6657
[root@node5 bin]# kill -9 6655 6657
-bash: kill: (6655) - 没有那个进程
-bash: kill: (6657) - 没有那个进程
[root@node5 bin]# service cloudstack-agent status
cloudstack-agent 已死,但 pid 文件仍存
[root@node5 bin]# rm /var/run/cloudstack-agent.pid
rm:是否删除普通文件 "/var/run/cloudstack-agent.pid"?y
[root@node5 bin]# service cloudstack-agent status
cloudstack-agent 已死,但是 subsys 被锁
[root@node5 bin]# service cloudstack-agent start
[root@node5 bin]# service cloudstack-agent status
cloudstack-agent (pid 109382) 正在运行...
[root@node5 bin]# netstat -antp |grep 8250
tcp    0   0 192.168.14.20:22220     192.168.14.10:8250     ESTABLISHED 109382/jsvc.exec 

处理后状态恢复正常,但是libvirtd仍然无法杀掉, 很快netstat -antp |grep 8250 状态再次消失,cloudstack master平台监控主机记录由Up状态转为disconnect状态。不过毕竟不是down状态,较之前已经有了进步。

启动一个libvirtd -d看下,

[root@node5 bin]# libvirtd -d
[root@node5 bin]# ps ax |grep libvirtd
  6485 "htmlcode">
[root@node5 bin]# libvirtd -d
[root@node5 bin]# ps ax |grep libvirtd
 130057 ?    Sl   0:38 libvirtd -d
 28904 pts/0   S+   0:00 grep libvirtd

至此既恢复了平台对该主机的管控,也终止了libvirtd异常进程。问题初步归于cloudstack-agent在处理发送个libvirtd的信号上存在些小问题。以后再单独分析下jsvc进程,再现问题和根本解决。

问题反思

在处理服务异常的问题上,命令行参数不要用restart,用stop和kill来调试。说起来都是泪!

上一篇:Centos 6.5中安装docker的步骤(简洁版)
下一篇:构建一个简单的CaaS系统
荣耀猎人回归!七大亮点看懂不只是轻薄本,更是游戏本的MagicBook Pro 16.
人们对于笔记本电脑有一个固有印象:要么轻薄但性能一般,要么性能强劲但笨重臃肿。然而,今年荣耀新推出的MagicBook Pro 16刷新了人们的认知——发布会上,荣耀宣布猎人游戏本正式回归,称其继承了荣耀 HUNTER 基因,并自信地为其打出“轻薄本,更是游戏本”的口号。
众所周知,寻求轻薄本的用户普遍更看重便携性、外观造型、静谧性和打字办公等用机体验,而寻求游戏本的用户则普遍更看重硬件配置、性能释放等硬核指标。把两个看似难以相干的产品融合到一起,我们不禁对它产生了强烈的好奇:作为代表荣耀猎人游戏本的跨界新物种,它究竟做了哪些平衡以兼顾不同人群的各类需求呢?