Start a Conversation

Unsolved

This post is more than 5 years old

1599

May 5th, 2009 02:00

SRDF/S DR scenario

In SRDF/S Disaster Recovery tests, what scenario should I choose. (DMX-3, mainframe environment, SRDF/S, no AUTOSWAP)

I would like to know any comments and suggestions about method of DR tests.

In documentation EMC recommends in SRDF/S (R1>R2)
following scenario:
1. make R1 NRDY
2. make R2 R/W
3. perform tests on R2 site (volumes)
4. optionally PREFRESH and PRE-RSUM changed tracks from R2 to R1
5. when test is finish, make R2 R/O, REFRESH R1 volumes from R2 volumes and finally resume ¿original¿ SRDF (RFR-RSUM)

In this scenario, application should be stopped (IPL-ed) twice and it is for me acceptable. During the test I haven¿t synchronization between R2 and R1 (I can only periodically PREFRESH and PRE-RSUM changed tracks). Advantage is that I can use any resynchronization method: form R2>R1 or R1>R2, full or tracks

In this case, why EMC don¿t recommend following scenario:
1. stop any I/O to volumes on R1 site
2. check that all volumes are in SYNC, no ADcopy, etc.
3. RDF-SUSP (R1>R2)
4. DELETEPAIR (R1>R2)
5. CREATEPAIR(NOCOPY) R2>R1
6. start tests and R/W activity on R2 site
7. when test is finish - stop application on R2 site and other I/O, make R2 R/O etc.,
8. RDF-SUSP (R2>R1)
9. DELETEPAIR (R2>R1)
10. CREATEPAIR(NOCOPY) R1>R2
11. start application on R1 site

In this scenario I take me own responsibility for consistency data during ¿NOCOPY¿ synchronization, but during tests (except two stops) I still have SRDF/S synchronization.

Any comments, pros and cons ?

Marcin

21 Posts

May 6th, 2009 01:00

Hi,

for testing R2, we use Autoswap (since 2006), but before we must use R2 and use EMC procedure only one time.

Just to understand what you want to do. If it's just to test R2 by stopping activity and restart On R1 after test or you stop activity on R1, restart by IPl-ing partitions, use R2 and after test restart production on R1 after refreshing from R2.

In anycase, it's not necessary to deletepair and re-create them. Just takecare when you restart SRDF. If you want to activate SRDF always R1>R2 you can use personalityswap on R1 and R2.


Deletepair is fast, but create pair is a bit slower and if you have many thousand dasd you save up time in your Test and eliminate risk on the appairage.

after stopping SRDF/S for a test, it's preferabale to relauch SRDF in AD mode for better response time especially if you have many invalid tracks and your applicatins are critical.

PS:

before using R2 in R/W you must change their status in ready mode before changing status in R/W.

have a nice test...

Message was edited by:
Notach

18 Posts

May 13th, 2009 00:00

Our DR-site was equipped with BCVs and we used these ones for DR-testing:

[1] Perform a Consistent Split
[2] IPL and run your DR-test
[3] RE-ESTABLISH the BCVs.

True, you'll need an extra set of volumes (BCVs) and of course, this is going to cost you. However, you'll be able to do DR-testing without interrupting your remotecopy. To us this was VERY important, important enough to get the extra volumes.

2 Posts

June 19th, 2009 07:00

If you need to do createpairs. There is an internal REXX program that can be used to call an API call instead of batch SRDF-HC. The API call is 10 to 20 times faster the batch SRDF-HC. I got the REXX program from my EMC Systems Engineer. The API did createpairs for 1200 volumes in 2 hours. Any chance you have SRDF/a ??? We use SRDF/a to take the consistent copy. Run testing on the R2's. Then restart SRDF/a using RNG-REFRESH , RNG-RSUM.
No Events found!

Top