09-16-2010 01:54 AM
I have four aix 6.1 vms which created by IBM P550 PowerVM. Two cluster my oracle database instance ,two cluster my sap central instance.
vcsdb: db1 db2 (database instance,virtual host name: dbv)
vcssap: ci1 ci2 (sap central instance,virtual host name:civ)
First I created basic resources for db and sap installation,such as dg,vol,mount,ip.
Then I should share some directory on civ for database instance installation, such as /sapmnt/PRD/exe /sapmnt/PRD/profile /sapmnt/PRD/global,So I configed NFS on ci1 and ci2 host,created share (/etc/exports).
Now,if virtual service is on ci1,then the share is available on ci1.If virtual service switch to ci2,then the share is available on ci2.
Then I have two questions:
1.When virtual service switch between two nodes, if should i restart nfs service?
2.after created nfs share,I added three mount resoureces to vcsdb cluster, such as civ:/sapmnt/PRD/profile civ:/sapmnt/PRD/exe civ:/sapmnt/PRD/global. But when i switch cluster to another node, the share could not mount properly,or failed. then you should restart nfs service on db1 and db2, then you can mount the share.
I hope to get any suggestion on this.
09-16-2010 02:06 AM
Have you configured a nfsrestart resource ?
Well as per experience, I will suggest restarting NFS is good & gives clean connections.... you can use NFSrestart resource to control this behaviour.... check out VCS Bundled agents guide, that will give you understanding on how to setup nfsrestart resource..
From Bundled agents guide
The NFSRestart agent recovers NFS record locks after sudden reboots or crashes on clients and servers. This avoids file corruption and provides high availability for NFS record locks.
The NFSRestart agent brings online, takes offline, and monitors the three daemons: smsyncd, statd, and lockd. If you have configured the NFSRestart agent for lock recovery, the NFSRestart agent starts the smsyncd daemon. The daemon copies the NFS information on the location of connections from the shared-storage to the local directory (/var/statmon/sm) and vice-versa.
Note: NFSv4root and NFSSecurity require AIX 5.3 ML03.
For important information about this agent, refer to: “NFSRestart agent notes” on page 122
This resource must be at the top of the resource dependency tree of a service group. Only one NFSRestart resource should be configured in a service group. The NFSRestart, NFS, and Share agents must be in same service group.
09-18-2010 08:39 AM
If you see the VCS Bundled agents guide
All the sample configurations mentioned for NFS do use NFSrestart resource... for e.g
Mount test_mnt (
MountPoint = "/test_mnt"
BlockDevice = "/dev/vx/dsk/test_dg/test_vol"
FSType = vxfs
MountOpt = rw
FsckOpt = "-y %-o full"
Mount test_lockinfo_mnt (
MountPoint = "/lockinfo"
BlockDevice = "/dev/vx/dsk/test_dg/test_lockinfo_vol"
FSType = vxfs
MountOpt = rw
FsckOpt = "-y"
NFS test_nfs (
Nservers = 20
NFSRestart test_nfsrestart (
NFSLockFailover = 1
LocksPathName = "/test_mnt"
NFSRes = test_nfs
Share test_share (
PathName = "/test_mnt"
Options = "-o rw"
Volume test_lockinfo_vol (
Volume = test_lockinfo_vol
DiskGroup = test_dg
test_nfsrestart requires test_ip
test_nfsrestart requires test_lockinfo_mnt
test_lockinfo_mnt requires test_lockinfo_vol
test_lockinfo_vol requires test_dg
test_ip requires test_share
test_share requires test_nfs
test_share requires test_mnt
test_mnt requires test_vol
test_vol requires test_dg
NFSrestart resource notes contain some important information (to use fully qualified hostnames) or notes if you are using NFSv4, worth reading..