Quorum issues
In general, it is recommended to disable quorum for geographically mirrored volume groups in order to minimize the possibility of leaving the resource group in the ERROR state. In PowerHA® SystemMirror®, quorum loss for a volume group causes PowerHA SystemMirror to selectively move the affected resource group to another node. Similarly, in GLVM for PowerHA SystemMirror Enterprise Edition, if quorum is enabled and then lost for a geographically mirrored volume group, PowerHA SystemMirror might start a selective fallover event (logged in hacmp.out as an rg_move for the resource group).
The following statements explain why it is preferable to have a quorum setting disabled. Depending on whether you have quorum enabled or disabled for a geographically mirrored volume group, these scenarios are possible:
- Quorum is enabled. If you have configured just one XD_data network only and it fails (or, if all of your XD_data networks fail), the RPV servers are no longer reachable and the RPV clients fail. This could lead to a quorum loss in some cases. PowerHA SystemMirror launches selective fallover of the resource group upon the loss of quorum, and attempts to move the resource group after the failure of the XD_data network, even though no other node at the site is active on the XD_data network. This leaves the resource group in the ERROR state. For information on how to manually recover the resource group, see No automatic recovery of the resource group in the error state.
- Quorum is disabled. If the XD_data network fails (or, if you have more than one XD_data network, then all of them fail) and quorum is disabled, it is more likely that access to the active volume group is maintained in the cluster. For instance, in a cluster with two nodes, one at each site, and two disk enclosures at each site, if the XD_data network fails, the access to the disks at the remote site is lost. But, if the resource group with the geographically mirrored volume group is hosted at another site (where disks are functioning), it can still remain online.
- Another reason to have quorum disabled is that it allows you to shut down the other site, for maintenance purposes. In this case, half of the disks are not accessible. But since quorum is disabled, and at least one disk remains available at the local site for the resource group, the resource group can still remain online on the node with access to that disk at the local site. As a result, when you need to perform a site maintenance, PowerHA SystemMirror does not attempt to move the resource group to any other node at the remote site.
To summarize, it is recommended to disable quorum for geographically mirrored volume groups and to have more than one data mirroring networks configured. This minimizes unnecessary fallovers and also allows GLVM for PowerHA SystemMirror Enterprise Edition to use surviving networks for data mirroring and potentially recover the resource group on the node at the remote site.