[Info-vax] VMware
Hans Bachner
hans at bachner.priv.at
Thu Dec 12 15:24:24 EST 2019
IanD schrieb am 12.12.2019 um 13:47:
> A point of interest, Charon I believe claim VMware certification
Yes, it does.
> Not sure if that includes an OpenVMS cluster or individual OpenVMS nodes migrating, I suspect the latter
I assume with "migrating" you mean moving a VM to a different host with
vMotion.
In the context of CHARON, "migration" is used for the process of moving
a physical Alpha/VAX to a CHARON instance.
But yes, using vMotion works just fine for CHARON hosts and the OpenVMS
systems running in them. As has been mentioned before in this thread,
you usually will observe a single ping showing higher latency, but the
systems stay up and running.
I did not yet try to move a VM with a single CHARON instance acting as a
cluster member, but I have a customer who runs a two-node cluster (with
a quorum disk) on a single VMware based CHARON host. You can move the VM
around between two datacenters ~10 km apart with no visible problems,
the cluster just keeps running.
My customers routinely use vMotion for both Windows and Linux based
CHARON hosts.
> [...]
>
> I think OpenVMS might need a bit of work fitting into a virtualized world especially where migrations are happening. The complexity in handling a node member of an OpenVMS cluster member might be very interesting indeed.
The OpenVMS cluster software is sufficiently tolerant to minor network
delays (if not agressively configured).
The only problem I have heard of, though several years ago with slower
hardware and networks, was a system running Rdb. If Rdb had to handle
serious load, its heavy use of memory caches led to the situation that
copying memory contents just wasn't fast enough to keep up with Rdb
activities and vMotion took forever.
> A single standalone node is one thing, migrating a cluster member might be a very different matter. You might have to come up with a formula not to migrate more than x members of an OpenVMS cluster at once etc. I don't know, I just know that when you start to deal with HA clusters things get exponentially complex very quickly
A VMware HA cluster in most cases won't help because it just reboots the
VM on a different host. I don't know how CHARON/OpenVMS work in an FT
configuration. VMware supported FT only for single vCPU VMs for a long
time, while CHARON required at least two (v)CPUs. I did not look at FT
configurations since FT supported VMs with multiple vCPUs.
> It's very good news that OpenVMS is looking to be supported in VMware, very good news indeed
+1
Hans.
More information about the Info-vax
mailing list