Vic/notes/Veritas

From Summerseas

Jump to: navigation, search

Contents

Veritas

CVM/CFS

  • How to determine which node is master.
    • vxclustadm -v nodestate
    • Or - vxdctl -c mode
  • Adding a CFS mount to the cluster
    • cfsmntadm add SkyLab1 SkyLab_vol1 /SkyLab1 SkyLabGroup1 all=cluster
  • Setting primary node for a CFS mount resource.
    • cfsmntadm setpolicy /SkyLab1 shsun07
  • Display info for a CFS mount resource.
    • cfsmntadm display -v /SkyLab1
  • Show primary node for a CFS mount.
    • fsclustadm -v showprimary /SkyLab1

VxVM 5 Notes

  • I/O Fencing
    • I/O fencing changed a bit in ver 5. There is now a fencing mode file, /etc/vxfenmode. Copy one of the files from /etc/vxfen.d to /etc/vxfenmode.
  • To add fencing to a VCS ver 5 cluster follow these steps...
    • 1. Copy one of the files from /etc/vxfen.d to /etc/vxfenmode.
    • 2. Create a 3 lun disk group named vxfendg.
    • 3. echo "vxfendg" > /etc/vxfendg
    • 4. /etc/init.d/vxfendg start
    • 5. gabconfig -a (Check that port b comes up)
  • Once fencing is configured, shutdown the cluster and add "UseFence = SCSI3" to main.cf in the cluster section right after the "Administrators = { admin }" line. Restart the cluster from the node with the updated main.cf and verify that this nodes builds locally.
  • Final check
    • vxdisk -o listreserve list (Any data disk imported by a cluster node should have reservation keys.)

Tuning

  • Fast Recovery
    • vxdmpadm settune dmp_fast_recovery=[on | off] (This is on by default)
    • vxdmpadm gettune all (Show tunables)
  • Restore Daemon and policy
    • /usr/sbin/vxdmpadm stop restore
    • /usr/sbin/vxdmpadm start restore interval=60 policy=check_disabled
    • Note-1 - The restore daemon may be started with the desired tuning at boot by editing the following file:
      • /lib/svc/method/vxvm-sysboot
    • Note-2 - To check the current setting...vxdmpadm stat restored
  • Setting the load balancing policy
    • vxdmpadm setattr enclosure <enclosurename> iopolicy=balanced
  • Setting the Log Level
    • vxdmpadm settune dmp_log_level=4

VCS 4.x and Solaris 10 NFS

  • The following is required on Solaris 10 prior to creating and NFS service group in the cluster.
    • svccfg delete -f svc:/network/nfs/server:default
    • svccfg delete -f svc:/network/nfs/mapid:default
  • Note - The services may be recreated easily by re-importing the service description files.

Repairing Volumes

  • Often while running fault injections a volume may become damaged. It is sometimes easier to repair the volume than to destroy and recreate it. This procedure is useful for volume repair.
    • vxprint -ht (Look for DISABLED plex)
    • vxmend -g ntap-test -o force off ntap-test1-01
    • vxmend -g ntap-test -o force on ntap-test1-01
    • vxmend -g ntap-test -o force fix clean ntap-test1-01
    • vxvol -g ntap-test startall
  • At this point the filesystem on the volume may be recreated or fcsk'd.

Setting the default disk group after install

  • vxdctl defaultdg dg_3020

Documentation

Troubleshooting

Long Boot Times

  • With VxSF5 and Sol 10 extremely long boot times may be experienced. Adding the following to /etc/system may resolve the problem: "exclude: drv/vxgms"
Personal tools