mcsg nfs toolkit

[nobleboi]

hv confired an nfs cluster using nfs toolkit...
am able to manually failover the package etc.
but its giving problem, lets say when i stop the nfs server service from the package running node.

[Geoff Wild · Accepted solution]

DO NOT STOP THE NFS service manually on a node in a cluster using NFS.

The only time you can safely do that is when all the packages are down first.

Rgds...Geoff

[nobleboi]

hi
is there a better way to test the package monitoring ?

[nobleboi]

its detecting the service failure and disabling the package switching (instead of node switching ?)

Jan 15 16:01:22 node1 syslog: Request from root on node node1 to modify package switching for package NFS_SERVER_PKG
Jan 15 16:01:22 node1 cmcld[20374]: Request from root on node node1 to modify package switching for package NFS_SERVER_PKG
Jan 15 16:01:36 node1 cmcld[20374]: Service NFS_SERVICE terminated due to an exit(1).
Jan 15 16:01:36 node1 cmcld[20374]: Service NFS_SERVICE in package NFS_SERVER_PKG has gone down.
Jan 15 16:01:36 node1 cmcld[20374]: Disabled node node1 from running package NFS_SERVER_PKG.
Jan 15 16:01:36 node1 cmcld[20374]: Failing package NFS_SERVER_PKG on node node1 due to service failure.
Jan 15 16:01:36 node1 cmcld[20374]: Request from node node1 to fail package NFS_SERVER_PKG on node node1.
Jan 15 16:01:36 node1 cmcld[20374]: Executing '/etc/cmcluster/pkg_nfs/pkg_nfs.cntl stop' for package NFS_SERVER_PKG, as service PKG*25089.
Jan 15 16:01:37 node1 syslog: cmmodnet -r -i 172.27.7.114 172.27.7.0
Jan 15 16:01:37 node1 syslog: cmhaltserv NFS_SERVICE
Jan 15 16:01:37 node1 LVM[1469]: vgchange -a n vg01
Jan 15 16:01:37 node1 cmcld[20374]: Service PKG*25089 terminated due to an exit(1).
Jan 15 16:01:37 node1 cmcld[20374]: Halted package NFS_SERVER_PKG on node node1.
Jan 15 16:01:37 node1 cmcld[20374]: Package NFS_SERVER_PKG halt script exited with NO_RESTART.
Jan 15 16:01:37 node1 cmcld[20374]: Examine the file /etc/cmcluster/pkg_nfs/pkg_nfs.cntl.log for more details.
Jan 15 16:01:37 node1 cmcld[20374]: Switching disabled on package NFS_SERVER_PKG.
Jan 15 16:01:37 node1 cmcld[20374]: Request from node node1 to disable global switching for package NFS_SERVER_PKG.

====

########### Node "node1": Halting package at Tue Jan 15 16:01:36 WAT
2008 ###########
Jan 15 16:01:36 - Node "node1": Unmounting filesystem on package_vip:/exports/d01
WARNING: Running fuser to remove anyone using the file system directly
.
package_vip:/exports/d01:
nfs umount: inform_server: package_vip:/exports/d01 server not responding:
RPC: Program not registered
Jan 15 16:03:29 - Node "node2": Unmounting filesystem on package_vip:/exports/d01
WARNING: Running fuser to remove anyone using the file system directly
.
package_vip:/exports/d01:
nfs umount: inform_server: package_vip:/exports/d01 server not responding:
RPC: Program not registered
Jan 15 16:01:36 - Node "node1": Remove IP address 172.27.7.114 from subnet 172.27.7.0
HANFS -- Jan 15 16:01:37
nfs unshare: /exports/d01: not shared
ERROR: Function nfs_un_export_fs
ERROR: Failed to unexport /exports/d01
HANFS -- Jan 15 16:01:37 - Node "node1": Halting NFS service NFS_SERVICE
cmhaltserv: Service name NFS_SERVICE is not running.
HANFS -- Jan 15 16:01:37 - Node "node1": Killing rpc.statd
HANFS -- Jan 15 16:01:37 - Node "node1": Killing rpc.lockd
HANFS -- Jan 15 16:01:37 - Node "node1": Restarting rpc.statd
HANFS -- Jan 15 16:01:37 - Node "node1": Restarting rpc.lockd
ERROR: Function verify_ha_nfs
ERROR: Failed to stop NFS
Jan 15 16:01:37 - Node "node1": Unmounting filesystem on /dev/vg01/lvol1
Jan 15 16:01:37 - Node "node1": Deactivating volume group vg01
Deactivated volume group in Exclusive Mode.
Volume group "vg01" has been successfully changed.

########### Node "node1": Package halt failed at Tue Jan 15 16:01:37
WAT 2008 ###########
node1:/etc/cmcluster/pkg_nfs#

[Geoff Wild]

NFS on HP-UX is very touchy.

Throw ServiceGuard into it - and anything that touches NFS outside of ServiceGuard can produce results that the only way to fix is a reboot.

Now, if all you want to do is test - as long as you arn't impacting production - ie you can reboot after your test, then just try killing processes manually.

Like this one will cause a failover:

/usr/sbin/rpc.pcnfsd

Kill that and the package should failover (as long as package switching is enabled)

cmmodpkg -e PACKAGENAME

I reiterate - you should reboot after your test - as NFS is very sensitive in a Service Guard environment.

Rgds...Geoff