redhat AS 4用RHCS做HA，断掉第一台机器网线，服务不能切换！

su_hub

仔细看一下/var/log/messages文件，你会找到答案的

su_hub

Dec 28 07:34:48 vm002 kernel: CMAN: removing node vm001 from the cluster : Missed too many heartbeats
Dec 28 07:34:48 vm002 fenced[2591]: vm001 not a cluster member after 0 sec post_fail_delay
Dec 28 07:34:48 vm002 fenced[2591]: fencing node "vm001"
Dec 28 07:34:48 vm002 fence_manual: Node vm001 needs to be reset before recovery can procede.  Waiting for vm001 to rejoin the cluster or for manual acknowledgement that it has been reset (i.e. fence_ack_manual -n vm001)

仔细看看红色的部分吧

su_hub

已经告诉你了

su_hub

Dec 28 07:34:48 vm002 fence_manual: Node vm001 needs to be reset before recovery can procede.  Waiting for vm001 to rejoin the cluster or for manual acknowledgement that it has been reset (i.e. fence_ack_manual -n vm001)

在vm002上执行
#fence_ack_manual -n vm001
因为你选择的fencing方法是手动嘛，你当然得“手动”告诉正常的节点有一个设备失效了，这样它才能“放心”的接替服务对吧？！

su_hub

那就需要有相应的fence device了，如APC等

su_hub

如果是使用RHCS3的话，也可以用quorum disk的，好像rhel4.4恢复了这个功能，不过没试过

su_hub

我记得RHCS手册上，添加fencing device这一部分有详细说明的，你的brocade switch好像也有包括，你可以去看一下

su_hub

Red Hat Cluster Suite 4 RHEL4 U4 Release Notes


Copyright(c) 2006 Red Hat, Inc.
        -------------------------------------------------------

September 19, 2006

Introduction

   The following topics are covered in this document:
   
     o Changes to Red Hat Cluster Suite 4

     o Important Notes
     
     o Bugs Fixed in the Release

     o Related Documentation
     
   
Changes to Red Hat Cluster Suite 4

  Quorum Disk
   
   Quorum Disk is a new feature available with this release. The
   Quorum Disk feature (also known as qdisk) allows you to configure
   arbitrary heuristics so that each cluster member can determine its
   fitness for participating in a cluster. The fitness information is
   communicated to other cluster members via a "quorum disk" residing
   on shared storage.
   
   With properly configured heuristics, you could define the following
   cluster behavior:

   * In the event of a network-partition failure, provide a method to
     decide which member wins the fence race in a two-node cluster.

   * Allow continued cluster operation after a majority failure without
     manual intervention.

  Quorum Disk communicates with CMAN, ccsd (the Cluster Configuration
  System daemon), and shared storage. It communicates with CMAN to
  advertise quorum-device availability. It communicates with ccsd to
  obtain configuration information. It communicates with shared storage to
  check and record states.

  You can find more information about Quorum Disk in the following man
  pages: mkqdisk(, qdiskd(, and qdisk(5).

  NOTE: For this release, you must configure Quorum Disk by editing
  the cluster configuration file, /etc/cluster/cluster.conf, directly
  rather than by using the cluster configuration graphical user
  interface (system-config-cluster).

ccs_tool Enhancements
  
  The ccs_tool includes new commands for this release. The new
  commands provide the ability to configure certain portions of the
  cluster configuration file (/etc/cluster/cluster.conf). In previous
  releases, the only tool available for creating and managing the
  cluster configuration file was the Cluster Configuration GUI
  (system-config-cluster). For more information about the new commands
  and usage examples, refer to the ccs_tool man page, ccs_tool(.


Important Notes

  The up2date command has changed for RHEL4 U4. When installing
  cluster suite software, use this syntax:

  up2date --installall=<channel-label>

su_hub

You can find more information about Quorum Disk in the following man
  pages: mkqdisk(, qdiskd(, and qdisk(5).