Kees, > are you testing disaster situations
We are a GDPS site, 3 production parallel sysplexes, 2 - 3 sites, PPRC and Hyperswap, with unplanned Hyperswap in test parallel sysplexes (two of them). A smaller part of production volumes is not mirrored (customer's decision). Because of GPDS we do test outages quite regularly in our test sysplexes, simulating an outage mostly through generation of messages. The real storage subsystem outages are quite rare, because we need the help of the vendor's technician (field engineer?). In the production there are regular annual disaster tests for a whole site, but not with power outage (lot of non-mainframe servers could have problems establishing normal work). > or did you really have a complete power outage on a storage device and was the device configured sufficiently redundant and if so, how could this happen? There was a UPS problem, it registered(?) anomaly in the electricity and itself and cut the electricity on the site (that was my understanding). I came later in and had to do only with the problems on the mainframe side. The simplex devices (that is, not mirrored) were not critical for the production (the applications could go on with the half of the capacity). The only problem was that applications hanged as result of these allocations. Luckily, it happened in the night, and as the electricity came, we had the access to all the volumes again - no problems. Well, as Ed Finnell wrote: "when the subsystem gets IML'd (takes forever under normal circumstances)". Normally, it is a primary GDPS site for one production sysplex, but at the time that was our secondary site, so we lost only the systems running there. Actually we had seen this behaviour a few years ago, as we had done a lot of basic GDPS functionality testing: I/O intensive job writing to two different volumes on two subsystems, one is powered off, you get only the IOS002A message, the job is (if I remember correctly) all the time "in" in SDSF DA, no CPU-time change, then we activate the second box, job writes on as nothing happened. At that time we thought it was a nice feature <g>. Zaromil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
