Brian_Westerman wrote:
Fish,
Do you have any special TCPIP settings on the hercules side
to go with changing to the OSA from LCS?
Nope. Just what's in our documentation.
I find that OSA seems to be able to FTP at 3 times the speed
of the old LCS,
Interesting. I hadn't really noticed that myself. Of course, I never bothered to time it either, so I'll have to take your word for it.
but unfortunately if I transfer more than a couple gigabytes,
it tends to get lost somehow and times out.
Also interesting! As a programmer, that sounds very much like a possible memory leak somewhere. I wonder whether the leak (if it is a leak! I'm just guessing that it is!) is in z/OS or Hercules?
Question: do you notice an increase in memory use in your host (Windows, Linux, etc) when this happens? On Windows, you can use Process Explorer. On Linux, I don't know what you would use. On z/OS, I don't know what you would use either!
[...]
I'm wondering if I have missed something in my definitions.
I doubt it.
I have offload turned off for the physical windows TCPIP
adapter, and in my zos TCPIP parms I also have nochecksumoffload
specified, but no other TCPIP config parms relating to the OSA
adapter, I just let them all default.
Which is fine. The checksum offload setting would only affect the speed of the transfer (not having it enabled slows things down a wee bit), but should not affect the memory leak issue (IF, that is, that is indeed what the problem really is! I'm only speculating (guessing) at this point!).
Although I suppose it's possible that the leak (if that's what it is!) could be in the checksum offloading handling logic of whoever's got the leak (host or guest).
In any case, this *is* somewhat interesting.
I'm not going to waste any of my own time/effort digging into this issue myself, but if it becomes a serious problem, I would report it (either to IBM or Hercules, whoever's at fault) so it can be investigated and fixed.
Maybe there is another way to transfer the dasd than using ADRDSSU
on both sides. I had heard that some people use CCKDDASD but have
not tried that.
I believe you mean CCKDDUMP/CCKDLOAD, not CCKDDASD. :)
Yes, there is the CCKDDUMP/CCKDLOAD(*) technique that Mike Schwab mentioned in his reply to you that can offload (dump) a dasd on z/OS to a very large dataset (which can then be xfer'ed to your Hercules host using whatever means desired) and then turned into a Hercules CCKD (or CCKD64) dasd using Hercules's CCKDLOAD utility, but I personally have very little knowledge of it or experience with it.
Hope that helps!
---------------------
(*) CCKDDUMP/CCKDLOAD for legacy 32-bit CCKD dasd, or CCKDDU64/CCKDLO64 for 64-bit CCKD64 dasd. I think there's a .zip file in our 'util' directory. I believe the late great Greg Smith originally wrote it, followed by me and then someone else who made (or tried to make!) changes needed to support 64-bit CCKD64 dasds, but that's about all I know about it. I've never actually ever tried to use it myself.
--
"Fish" (David B. Trout)
Software Development Laboratories
mail: fish@...
