Hello braintrust,
My ovirt cluster is prone to hanging when it is put under a high IO load.
In an earlier thread, someone determined that this is consistent across
architectures and infrastructures when NFS is used for the backing
datastore for the cluster.
I was able to see a huge improvement by switching putting an nVME drive in
as the intent log cache on the NAS. But, alas, that appears to have only
put off the reckoning, and the cluster locked up again when I doubled the
IO workload.
Prior email chains to this list have established that the hang occurs
because "something" in qemu throws the VM into PAUSE state instead of
IOWAIT state when under load. Other emails I have viewed seem to point at
NFS as the root cause, so a "least cost" suggestion is to migrate from NFS
to iSCSI as backing storage.
Is there an existing procedure for migrating Ovirt VM storage from NFS to
iSCSI?
Current architecture:
Ovirt engine 4.4.4.7-1.el8 on centos 8.x, on bare metal intel desktop
2 hosts 4.4.4.7-1.el8 on centos 8.x
TrueNAS Core 2020, with shares exposed as NFS
10 GBit dedicated storage Network
Thanks,
*David Johnson*
