[Info-vax] free shell accounts?

[David Froble]

Simon Clubley wrote:
> On 2015-01-22, David Froble <davef at tsoft-inc.com> wrote:
>> Stan Radford wrote:
>>> I don't understand what a cluster does. If they don't have shared disks
>>> somewhere wouldn't they have to have multiple copies of everything? How does
>>> a cluster still remain usable if you are editing a file and the machine the
>>> file lives on fails? I can see for serving applications a cluster would be
>>> great but I don't understand how it helps development users. And even that
>>> would seem like it would take a lot of planning and wouldn't just automatically
>>> "work" because of the need for shared storage somewhere.
>>>
>> The concept of a VMS cluster is "shared everything".  There can be 
>> various implementations.
>>
>> For best versitility, each computer (node) would have a connection to 
>> every device.  Thus, if a particular node in the cluster goes down, all 
>> other nodes still have access to all devices.
>>
> 
> I wonder if what Stan is missing is the concept of the DLM and what that
> enables.
> 
> Stan, in a shared everything configuration, the disks do not belong to
> a specific machine but all VMS machines in the cluster can all write to
> the same disk (and update the filesystem structures on that disk) at the
> same time.
> 
> This is made possible in VMS land by the use of a Distributed Lock Manager
> (or DLM for short). The DLM is a cluster wide common database of locks
> and allows multiple VMS systems physical access to the same disks along a
> common hardware path as the DLM allows these multiple VMS systems to
> automatically coordinate their activities with each other.

Yes, this is what makes VMS clusters with shared everything possible.

> This includes a machine leaving (or joining) the cluster which is also
> handled automatically by an automatic procedure known as a cluster state
> transition. Once the state transition is complete, the remaining nodes
> just carry on from where they were.
> 
>> Now, it is also possible for devices to be connected to a single node, 
>> and that node can share them with the cluster, but such devices are lost 
>> to the cluster if that node goes down.
>>
>> With today's technology, some people use SANs, and then each node in the 
>> cluster is basically a "compute device".  As long as one node and the 
>> SAN is up, the cluster exists.
>>
>> I may not be the best source of cluster information, and I'm pretty sure 
>> that except for an brief overall description a newsgroup post is not the 
>> best place to learn about cluster details.
> 
> I would agree with the latter bit about learning cluster details. :-)
> I think Stan has some reading ahead. :-)
> 
> Simon.
> 

Yes, some reading might be in order, if he really cares about the 
details.  But one can use a cluster without being buried by all the details.

One thing I left out and Philip mentioned, is shadowing.

Simple case:

Node A with one disk
Node B with one disk

HBVS (host based volume shadowing) is used to make both of the disks a 
single shadow set.  Thus, what's on one disk is also on the other disk.

Node B leaves the cluster.  Node A still sees what is on the disk.