[Info-vax] free shell accounts?

[Stephen Hoffman]

On 2015-01-23 12:51:40 +0000, Stan Radford said:

> On 2015-01-22, Phillip Helbig (undress to reply) 
> <helbig at asclothestro.multivax.de> wrote:
>> In article <m9r1cq$p2q$1 at speranza.aioe.org>, Stan Radford
>> <sradford at noemail.net> writes:
>> 
>>> As long as they're all physically attached to the same drives. But if
>>> they're attached to network drives and the owning box goes down then there's
>>> a problem.
>> 
>> Not if the disks the applications see consist of more than one physical
>> disk and/or connected to more than one node.
> 
> If the file you want is on the portion of a volume that's only partially
> available now that the only owning box went down and took that part of the
> volume with it then clustering and the fact the rest of the volume might be
> available doesn't help you very much until the part you want comes
> online. However it is interesting that the rest of the volume or disk group
> or whatever it's called in VMS is still around.

OpenVMS is not Unix.

OpenVMS Clustering doesn't work the way that you seem to think it does.

With host-based volume shadowing (host-based RAID-1) in a cluster, 
there are no partial volumes.  All of the data on the shadowset is 
available from all of the operational members in the cluster, up until 
the last member volume goes offline for whatever reason.   It's 
host-based RAID-1, after all.  With current VMS, six separate member 
volumes are permitted in the shadowset, and some or all of these member 
volumes can themselves be hardware RAID volumes.  So assuming a cluster 
with three lobes across a 100 kilometer metropolitan area and with two 
volumes located in each lobe of that cluster, and with each of those 
six member volumes then RAID'd across some number of physical disks, 
getting a shadowset volume to go offline takes a whole lot of failures. 
  From the perspective of application software, the shadowset — also 
sometimes called the DS disk or the virtual unit — is simply a regular 
storage volume.  Barring caching hardware that lies to the host, when 
the write is completed back to the application, the write has been 
performed across all member volumes.  Downside?  HBVS has to get the 
I/O data to all volumes synchronously, and that takes some time.

There are mechanisms to speed adding volumes back into a shadowset and 
this can be useful for various purposes, and there's a whole lot more 
detail on HBVS in... wait for it... the HBVS manual.






-- 
Pure Personal Opinion | HoffmanLabs LLC