>I am curious if you have tried having your 'keep-it-open' program >close and re-open the file every 10 min? Or maybe try having the >program issue a DESERV to read the directory every 10 min. Seems >silly to have to do something like this, but maybe it would get >the system to keep the directory cached.
No, I didn't. close and open itself doesn't really help, it just keeps the 'connection' to the PDSE. *Someone* has to go and really read the directory to get it into cache, and a sequential reading of the directory for all of the affected datasets here takes about 45 minutes, way more than the 15 minutes maximum the caching effect lasts, so I didn't bother. Remember, I have about 26 of those big beasties. For what it's worth, *right now* buffer_beyond_close appears to work. But we've had that before, and then at some point it stopped again. So I am just waiting and this time around will pounce when it doesn't work. The task is to take a dump about 5 minutes after the closing of the dataset. And I was told that in my case the VB format of the members is the reason for the slow reading of the directory, as apparently for VB there is a lot more I/O going on. It would explain why I saw more than 10000 I/Os for the job when I 'only' had 10000 member in it. The jury's out on that, but at least I now have an ETR open for this. (Plus my other guesses about what happens were more or less right, not that that is any consolation.) Best regards, Barbara ---------------------------------------------------------------------- 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
