VCS无法failover 数据库(fsck太久）的解决办法

Document ID: 252030
http://support.veritas.com/docs/252030 E-Mail this document to a colleague
UNIX File System logging minimizes fail-over time for a service group containing a mount resource that is a UFS file system.
--------------------------------------------------------------------------------
Details:
In most cases when a system crashes, file systems will need to run fsck before being mounted again.  In the case where a VERITAS Cluster (VCS) service group has a UFS file system configured within it, the fsck during the online of a group after failure could exceed the amount of time allowable for a resource to online.  This could result in VCS being unable to online the service group in the cluster after a system crash.  

In Solaris 8, Sun introduced logging for UFS file systems.  In most cases, fsck done on a UFS file system with logging will use the logs to recover from a file system fault.  This could save a substantial amount of time during fsck.  The fsck portion of the online for the mount resource would be able to finish before VCS considered it "timed out", unless of course a full fsck is still needed to be done on the file system.  

To enable the UFS logging option for a VCS mount resource that is a UFS file system,  do the following:

#/opt/VRTSvcs/bin/haconf -makerw
#hares -modify <mount_res_name> MountOpt logging
#haconf -dump -makero

Increasing the OnlineTimeout limit for the mount agent to give fsck more time to complete can also be considered in addition to using UFS logging.  This could be helpful in the case where a full fsck still needs to be done on the file system.  This will change the OnlineTimeout value for all mount resources.  The default value is 300 seconds.  To increase the OnlineTimeout value do the following:

#haconf -makerw
#hatype -modify Mount OnlineTimeout 600
#haconf -dump -makero