11-03-2010 07:31 AM
On Mainframe Configuring 4 DCX Ficon Directors and none of the connections get physical login.
IBM z/10 shows invalid attachement failure for all channels on HMC of z/10. Attached devices as well
(EMC VMA, IBM VTS). Looks like a fundamental setup problem but can't be found.
Can you maybe look at the inital setup to maybe see the obvious. We will provide config details as
requested. Many thanks.
11-03-2010 12:11 PM
On DCX follow Ficon Release are Supported and Certified:
v6.3.0b is Generally not Supported for Ficon.
If you are satisfied with the responses please mark as 'Answered'.
If not, just let me know I can see to get more info.
I would appreciated if you rate the threads.
11-03-2010 04:21 PM
As this is obviously a new implementation, it would be recommended that you upgrade the DCX's to 6.4.0c
You also mention 4 x DCX, so I assume you have 2 fabrics each with 2 x DCX, and therefore will be 'cascading'
i.e. the z10 CHPID attached to one DCX with the Target CU on a different DCX ?
To accomplish this you must have a 'High Integrity Fabric' and the z10 CHPID will test for this being in place.
If the CHPID does not find the required fabric settings it will not come online and you will see 'Invalid attachment'.
To implement a High Integrity Fabric you must:
Turn on the insistent domain ID flag (IDID) on all switches.
Enable port-based routing on 4 Gbps and above switches pre FOS 6.4.
Disable Dynamic Load Sharing (DLS) on all switches pre FOS 6.4.
Enable In-Order Delivery (IOD) on all switches.
Set High Integrity Fabric Configuration (HIFC) on the seed switch:
Fabric-wide consistency policy is configured to include SCC in strict mode.
SCC policy is created or modified to limit connectivity to only the switches in the selected fabric.
(Optional) Turns on FICON Management Server (FMS) mode on all switches.
11-04-2010 12:15 AM
I saw this behavior a few weeks ago as well during a installation.
In my case the Fabric-Wide Consistency policy was not set. Ones I created it, the port of the mainframe logged in.
You can verify with the fddcfg --show command. If the policy doesn't exist you can create it with the fddcfg --fabwideset "SCC" command.
11-04-2010 07:13 AM
here is the solution. Initial problem is that the DCX has been configured after the z/10 CPU was POWER ON RESET.
No Host Channel logged in after the channels were selectively toggled off and on.
The sequence to get all channels logged in is important. This is different with former i10K Brocade and ED64 McData directors.
1. Configure the Ficon Directors (DCX) completely.
2. Power on Reset z/10 with corresponding IODF.
Now the Security Fabric Feature will be activated in the Directors.
The DCX will establish the Connection Control on request of the z/10.
When the IODF is loaded before the DCX is configured (we have done this) you will never get it running.
A single Ficon Card Toggle on z/10 can never initiate a login on a Port on a DCX.
A toggle of the Channels for a single LPAR will not change the status unless all attached LPARS
This were our first experiences with DCX and it looks like this is a significant change to former described directors / Switches.
Solution has been provided by IBM Technical Support in Mainz when opening a problem referring to our INVALID ATTACHEMENT Failure.
Many thanks to all who tried to help me and gave advise.
01-21-2012 04:32 AM
I've encountered the same problem in a new installation.
Where can I find info about which FOS versions for model 5300 switches support ficon? We currently run 6.4.2a.
I will also check the Fabric Wide Consistency policy first thing on Monday morning thanks @Macguyver