Disaster Recovery Config

[Olebile Molefi_1]

Gentlemen,

Please kindly throw-in your expert advice on the best practice standard for configuring 24/7/365 Production Site (Single Domain) vs DR Site over a 10km fibre connection?

[Jan van den Ende]

Olebile.



I do not fully understand 24/7/365.
One way to say "uninterrupted" is 24 by 7,
meaning every hour of the week.
24 by 365 means about the same, but much more emphasis on keeping it up for longer stretches. (and purists say 7 by 365.25, to include the quarter-day-per-year that gives the leapday every 4 years)
But 24/7/365 looks like 365 weeks, being nearly 7 years...



First question:
Are you locked to any specific OS, or do you still have a choice in it?

A lot of your questions are covered by
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=132956
(but you probably found that allready)

And if you are not already locked into any of the *NIXes, have a look at
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=550510

If the intent is to do 24*365.25 using M$ toys, then: DON'T get burned, just RUN AWAY!

And DO make your plans first then thoroughly (have them) evaluate(d) in the most pessimistic perspective, then move ahead, and KEEP re-evaluating constantly.
As we proved, it CAN be done!

Success!

and in a few years, publish YOUR story here!


Jan

[tazei]

very good !

[Tor-Arne Nostdal]

We didn't get our second computer room, so we have chosen a solution based on:

1)
MC Serviceguard clusters in our single computer room. This ensure high availability to our DB-servers.
Auto switchover take approx. 15min. in case of system breakdown.

2)
Emergency reports scheduled each night and distributed to local sites.
This ensure organisation to be able to use manual procedures in case of total breakdown of entire computerroom (bomb/fire)

3)
Agreement with external company to host our system (in another country...).
We continously keep them updated with our system configuration.
Backup tapes are stored in locked suitcases and stored externally - and if disaster occure they're shipped to SP (extra key pair at external SP).
We take extra archive/redolog backups out twice a day.

If a full disaster occure we're up-and-running within 72hours.

We test the solution once a year.

/Tor-Arne

[Chuck Ciesinski]

Olebile,

A lot of what you need to get answers for depends on what type of application you are running, how much risk your management wants to take and how much money they are willing to spend. Jan has provided you with several good links. One key factor to consider is, 'What type of disaster am I protecting against?'. Disasters can be internal or externally caused.

If you are running HP-UX or Tru-64, you are already on a good start. We have 5 different clusters, applications based, with MCServiceGuard and are very happy with the results. I cannot speak for Tru-64, but I don't think HP would have bought an O/S unless there was a lot of good about it.

Remember, it's not just about the cpu and data, it's also the datacomm, getting users on-line, printing, getting trained and qualified staff to the alternate disaster recovery site, and having GOOD documented proceedures to recover
from.

HTH,

Chuck Ciesinski

[Jan van den Ende]

Olebile,

One guy doing A LOT in this area is Keith Parris. If you check him out at
http://forums1.itrc.hp.com/service/forums/publicProfile.do?userId=CA1001515&forumId=1
(watch out for the wrapping URL!)
in his postings you will find approx half of them dealing with this issue.

And I would like to repeat here his favorite statement:

" The best disaster recovery is NO recovery, but PREVENTION". Think about it...

Success.

Jan