different TDF in a cluster?

[Ian Miller.]

Is it possible to run nodes with diffent time zones in a cluster. The nodes will have their own system disk (as they are at diffent sites).

[labadie_1]

I do not have a direct response to this problem. But if memory serves me, there used to be a tool that allowed different users to use different times. So a user USER1 could send batches with a time of 2 years ago, a USER2 could use a time of next year...

It was a hack (what a surprise !) :-)

and needed some more code to work with Vms 7.3+

[labadie_1]

and some tools like Cyrano (appeared for W2K testing in 1999 or before) "play" with the time. That may suit your needs.

[Martin P.J. Zinser]

One such tool is datesim from software partners

http://www.sp32.com/

Besides of the general confusion a cluster in several timezones is sure to create (have fun with submit/after to generic queues ;-) Problem might also be cluster communication if the timestamps of messages differ to wildly.

Greetings, Martin

[John Gillings]

Ian,

There is no supported way to operate a cluster with any substantial difference in times between nodes. Odd things happen!

Although there are numerous tricks and hacks which can give the appearance of different times or time zones between nodes or processes, there are architectural reasons why these won't work in 100% of cases. Unfortunately "time" under OpenVMS wasn't abstracted enough in the original design, so there isn't a single place to catch requests. Dates and times are distributed too far and wide. For example, consider date stamps on every file. You'd need to know the zone of the writer in order to translate it into the zone of the reader.

Individual processes can explicitly manipulate times using the various time services, but there is no way to (safely) do it transparently.

(ideally, any OS will keep track of, and store all times in UTC. It's then possible for each process to impose its own timezone when displaying the time. OpenVMS doesn't do this, so we're stuck with the single time zone per node model. Changing this would be an horrendous project).

[David B Sneddon]

Ian,

We use DATESIM (from Software Partners) although we are not using it in a cluster. The "system" runs at GMT and each "user" has their own timezone defined. Because of the separation of user and system activity this would continue to work within a cluster (for our situation). If you run all the systems at GMT (or at least in the same timezone) and have the user activity the only thing that is fiddled with, then it should work -- it would depend entirely on the software you are using and the types of user activity involved.
(I am not affiliated in any way with Software Partners or the creator of DATESIM)

Regards
Dave

[Lokesh_2]

Hi Ian,

If your systems are at differen locations, and not sharing system and data storage, then why you want them to be in cluster ? Why not separate system networked together...and using NFS, DFS etc..

Thanks & regards,
Lokesh Jain

[Ian Miller.]

thank you for your thoughts. As its not supported then I will have to think about non-clustered configurations or keeping the cluster on the same timezone.