We recently implemented SRDF/A with our 2 DMX4s between CA and TX. This first implementation is for our 2 z/OS mainframes one in TX and one in CA. During normal processing we have no problem keeping 30 second cycles. During heavy batch processing our cycles can elongate to 60 minutes or more. When we get too far behind SRDF uses too much cache and sometimes we drop. I am trying to use ECC to monitor SRDF. There is an alarm set up to send an alert if we get a certain wp_limit drop. We got a wp_limit drop during month end but did not get the aleret. It turns out there is a second condition that can trigger a slightly different wp_limit drop and it has its own alert that needs to be set. I'm concerned that some day we will drop with some new condition that is not caught by our currenn ECC alerts.
Is there a better way of monitoring SRDF/A, maybe outside of ECC. Maybe the mainframe could check something to make sure it is still up and alert if it is down. I am an Open Systems (Solutions Enabler etc) but I have become familiar with some of the mainframe SQ Query commands etc.
Joe, our systems programmer has been able to see these errors in the MVS logs so we can create an alert off them. We use a scheduling product called ESP and we my use it to kick off MVS jobs to restart SRDF/A. We might do something like this -
SC VOL,LCL(5000,03),ADCOPY-DISK,ALL
SC VOL,LCL(5000,03),RDF-RSUM,ALL
SC VOL,RMT(5000,03),RNG-REFRESH,ALL
SC VOL,RMT(5000,03)RNG-RSUM,ALL
SC VOL,LCL(5000,03),ACT
It seems like whenever SRDF/A drops the the 03 Mainframe group - the volumes go to TNR-SY (Sync) but if I do an symrdf query from open systems Solutions Enabler - it says it is in async mode. I don't know which system is giving us the true answer. We always run the SC VOL,LCL(5000,03),ADCOPY-DISK,ALL just to make sure we don't accidently restart in Sync mode as that would kill our performance.
What do you think about the 5 SC commands above for restarting. I know I can usually restart an open systems group simply by doing "syrmrdf resume". In some cases I may need to do a "symrdf est" (incremental establish). I want to test this out on the mainframe group at some point. If it works it would allow our Open Systems people to restart the mainframe SRDF/A group.
I've noticed after a drop while running the RDF_RSUM we always seem to get "Special Processing is Required". At that point we run the RNG-REFRESH and RNG-RSUM. Hopefully its ok to run both these jobs even if we don't get the "Special Processing is Required". As we want to set up 1 routine to always use after a drop. Sound ok?
Those 5 SC commands are appropriate. You may find that the RNG-REFRESH and RNG-RSUM are only needed if the RDF-RSUM fails with "devices are inelible for the action".
But always first check your operational mode and issue the ADCOPY-DISK,ALL... as you noted "sync mode" would not be good for production performance.
After sleeping on your questions, I think it would be best just to remove the RDF-RSUM all together since you are using the RNG-REFRESH and RNG-RSUM.
Doing both is redundant work and unnecessary. This would speed things up for you and also be a cleaner process.
Say by chance the RDF-RSUM did not fail, then the RNG- commands would not be necessary and would fail since you were already resumed. So, by removing the RDF-RSUM you should not have any commands to fail and this would be cleaner and less confusion.
dynamox
11 Legend
•
20419 Posts
•
87439 Points
1169
0
Posted November 8th, 2010 20:00
take a look at symrecover command