Implementing disaster recovery for HP Integrity Virtual Machines with Metrocluster and Continentalclusters on HP-UX 11i

18
Troubleshooting Metrocluster VM problems
This section describes the various problems that you can face while using Integrity Virtual Machines.
Each problem is followed by the associated workaround or solution.
If the virtual machine does not start up when you try to start the package manually
Try to start up the VM outside the Serviceguard environment, and see if it can be started properly.
1. For Metrocluster Continuous Access XP/P9000, check to make sure that the device group is in a
pair state using the
pairdisplay command.
For Metrocluster SRDF, check to make sure that the devices in the SRDF device group dgVM are in
a synchronized state with the “
symrdf –g dgVM query” command.
Similarly for Metrocluster Continuous Access EVA, check with Command View EVA to make sure
that the DR group operational state is good.
2. Use the hpvmmodify command to set the guest not to be distributed.
# hpvmmodify –P vmmetro –i NONE
# hpvmmodify –P vmmetro –j 0
3. Set up guest storage by importing the VxVM disk group vmdatadg.
# vxdg import vmdatadg
4. Start the logical volume.
# vxvol –g vmdatadg startall
5. Start the VM.
# hpvmstart –P vmmetro
6. Try connecting to the VM to verify that it is running properly.
# ssh vmmetro
If the VM is running properly now, re-create the Serviceguard package and try again to start the VM
as a Serviceguard package.
If the virtual machine does not fail-over to the remote site
1. Check to see if the VM storage is configured properly in the package control script.
2. Check the package log file. If the log file indicates that the Symmetrix/XP/EVA device is in a state
that does not enable automatic startup, then:
a) Fix the device state and restart the package.
b) Create the
FORCEFLAG file in the package directory or modify the AUTO* variables (for MC/
Continuous Access XP/P9000 or MC/SRDF) in the environment file, and restart the package.
Please note that forcing package startup this way might result in the package starting up on
non-current or inconsistent data.
3. Check to see if the correct device group is configured to be managed by Metrocluster.
Once you have completed the above procedure, the VM will fail-over to the remote site. If the VM still
does not fail-over, then contact your HP representative.
Application startup and monitoring
Because Metrocluster/Serviceguard can only control the startup of the VM guest in the “VMs as
Serviceguard packages” model, any applications within the guest must be started through the guest’s
defined initialization sequence, or by using the guest’s native operating system commands.