XCF signaling stopped working with ga36e0bbc

[edtgek]

Hi. After installing Hercules pre-built version Hercules-4.0.0.8873-SDL-ga36e0bbc, XCF signaling stopped working in both my base sysplex.

With version Hercules-4.0.0.8845-SDL-g15c01ddb and some earlier versions both my base sysplex is working great.

I tested the newest version a couple of times and the first system in the plex is coming up fine but the second system get no contact with the first system. This happens every time I test. You can see the response in the last picture in the attached file. As soon I go back to the version Hercules-4.0.0.8845-SDL-g15c01ddb both my plex works fine again.

In the attached file I show how the CTCE's are configured in the two Hercules configuration files as well as the corresponding parmlib couple members for both systems.

XCF_signalling_stopped_working.docx

[Fish-Git]

@edtgek : I am not familiar with sysplex and admit I have never tried to create on using CTCE.

I would like to try and recreate the problem for myself so that I can properly debug this issue.

What are the requirements? Do I simply define two Herc systems using cross-connected CTCE devices like you have, along with PARMLIB(COUPLE) members (with swapped PATHIN and PATHOUT statements) like you have? Is that all there is to it?

Thanks.

(And sorry about the breakage! With your help I'm sure I'll get it fixed!)

[edtgek]

Hi. You can try to test. You will need 2 CTC addresses in MVS that correspond with the CTCE in the Herc configuration. In `PARMLIB` in your `IEASYSxx` member, you will need the` PLEXCFG` parameter to be set to `MULTISYSTEM`. You also will need some couple datasets that you see in the couple member. In the two Herc configurations, the `LPARNAME` and `LPARNUM` should be unique -- _especially_ LPARNUM (e.g. `LPARNUM 1` for the first system and `LPARNUM 2` for the second). There may be more things that also need to be done, but as a first attempt, try the above and see if you can bring up each system, paying special attention to see what happens when the the second system tries to come up. Göran Eklöf Sjödalsvägen 26 141 47 Huddinge Sweden

[Fish-Git]

You will need 2 CTC addresses in MVS that correspond with the CTCE in the Herc configuration.

I presume you mean I should use CTC devices in each system's TCPIP PROFILE, each defined with a different IP address, yes? (i.e. a total of 4 different IP addresses? Two for each system, one IP address for each CTC device?) Is that correct?

You also will need some couple datasets that you see in the couple member.

How do I create these datasets?

Note!   I am very inexperienced with MVS/ZOS, so I will need lots of help from you! Thanks!

[edtgek]

No, TCPIP is not involved. To look at this problem I believe you will not even have to start any JES. You will need an IODF file which is the configuration of MVS, which devices the MVS system should be aware of and there you will need to have 2 CTC devices. It should be 2 CTC devices known to MVS and as defined in the couple members also known to XCF. In my couple members I am using E24 and E25 they are known to MVS and XCF and they are also defined in the HERC configuration file as CTCE as I showed in the file first attached. A sysplex is 2 or more MVS system that work closely together and they are using couple datasets to communicate with each other. I think your couple members can look like below, I think it is the only couple dataset you will need. I used this JCL to create the couple dataset.

[Fish-Git]

@edtgek : Göran,

It sounds like trying to recreate this problem might be beyond my abilities, so I would like to do something else instead.

I would like you to try my test Hercules containing my proposed (possible) fix. Please send me your email address. My email address is fish at softdevlabs dot com.

Thanks!

[Fish-Git]

Göran (and any other interested parties):

SDL Hyperion commit 71d9a74 fixes this problem.

Those building Hercules for themselves will simply need to do a git pull to pick up the change and then rebuild Hercules.

Those using pre-built Windows binaries will unfortunately have to wait for the next snapshot, which is more than likely still several weeks away yet.