[Info-vax] Looking to remote cluster my two OVMS machines. Questions. Need help!!!

[Stephen Hoffman]

On 2012-10-10 01:03:56 +0000, Len Whitwer said:

> I have two ES40 Alpha servers running OVMS 7.1 (75 miles apart)
> Machines are identical for disaster recovery. (Using tape now)
> Using Decnet phase IV. (Not DECnet over IP) I have COMCAST between
> the two sites with speeds over 15mb. Looking to keep costs down!!
> Questions!!!


> 1.  Has anyone already done this??

Clustering ancient software versions with big and slow and power-hungry 
boxes over potentially sketchy networks?   Sure.

> 2.  What networking equipment will support DECnet. (Routing or bridging)
>     o Have heard Cisco and Procurve might work.

If you don't need DECnet, then don't use DECnet.  Clustering doesn't 
require it.  If you need DECnet, then you can downgrade Phase IV to 
Phase V and use DECnet over IP.  Or yes, you can bridge Phase IV, given 
appropriate switches with network-bridging capabilities.

But I might guess that you are assuming that clustering is related to 
DECnet.  It's not.  Completely separate protocols.

As for connecting these hosts in a cluster, you can either bridge the 
connections and run SCS over IP and yes, that's based on the bridging 
capabilities of the switches and the latency and bandwidth of the link, 
or you can upgrade both systems to OpenVMS V8.4 and use SCS over IP 
(without needing the switches), though that too is dependent on the 
latency and bandwidth.

> 3.  Open to any thoughts??

Quit messing around with old hardware, ancient software, and install 
some more appropriate gear?

Or roll your backups out to a storage provider, and skip the need for 
multiple-site clustering?

But since you're likely going to persist in using the old stuff, well, 
sure, you can cluster this stuff.  Start with the clustering manuals as 
an introduction.  There are two clustering manuals in the shelf, a 
higher-level intro, and a lower-level manual with more details.  The HP 
documentation is available at <http://www.hp.com/go/openvms/doc>  Also 
see <http://labs.hoffmanlabs.com/node/153> 
<http://labs.hoffmanlabs.com/node/569> 
<http://labs.hoffmanlabs.com/node/349> and probably some others, for 
cluster and quorum and related details.

Or bring on somebody that knows how to do this.  The basic cluster 
setup is pretty quick to establish, though there's usually some 
troubleshooting involved somewhere.  The data link - if it's really 
delivering fifteen megabits, and assuming you have static IP on both 
ends - will be problematic given you're probably going to be thinking 
about using host-based volume shadowing - HBVS, disk mirroring - as 
that'll tend to saturate that link for disks of any size.  Officially, 
any link faster than a 10 megabit Ethernet is good enough.  But in 
practice, even boxes as slow as those two AlphaServer ES40 boxes can 
saturate that link all week long, particularly when HBVS is afoot.

What's your failover time requirement?  Applications that need 
continuous access?  Or can there be outages?  (The two-node cluster 
configuration isn't continuous access, and can't be.  It's always 
primary-secondary.)

But it'd probably be easier to upload backups of the data or the 
databases to a storage provider, and download and load the savesets 
when you need to restart processing.  Particularly if you don't need 
continuous access.



-- 
Pure Personal Opinion | HoffmanLabs LLC