-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Description
ISSUE TYPE
Bug Report
COMPONENT NAME
HA, Storage
CLOUDSTACK VERSION
4.11.0
CONFIGURATION
Basic networking, 3 NICs, management, Guest, storage, zone wide NFS storage
OS /ENVIRONMENT
CentOS Linux release 7.5.1804 (Core)
SUMMARY
When using multiple NICs as described above VM HA does not work if you have configured zone wide NFS.
When the power is disconnected from one of the compute nodes the management server logs show the ping timeouts but it never confirms that the host is down so in the UI its state is "Alert" not "Down". Because of this there is no attempt to restart the VMs on another of the compute nodes.
With cluster wide NFS instead VM HA works with multiple NICs.
The same happens with 2 as opposed to 3 NICs so I think it is just if you split traffic across the NICs.
I ran two comparison tests -
- all traffic over single NIC, zone wide NFS - worked
- management/storage on one NIC, guest on another, zone wide NFS - did not work
and I have the management server logs for both tests if needed ?