Sun StorageTek VSM4 System
 
Sun StorageTek VSM5 System
 

PLA-Support>Sun Systems>TAPE>Virtual Tape>SN-TP: VSM
 

In this Document

Symptoms

Cause

Solution

Created from <SR 3-6672327581>

Applies to:

Symptoms

NDCL (Non-Disruptive Code Load) is performed from microcode code D02.14.51.00/H02.14.50.00 or lower to microcode D02.15.20.00/H02.15.20.00 or higher.
Interface card is taken offline and a GFR (Guided FRU Replacement) NO-replace of the card is done to upgrade the code on the interface card.
When an attempt is made to put the interface back online it fails with a NED (Node Element Discriptor) problem.  Sample message:

13070 11:42:07.33 00000090 IOS291I CONFIGURATION DATA COULD NOT BE READ ON PATH(21C0,6F) RC=21 064
              064 00000090 TOKEN NED MISMATCH HAS BEEN DETECTED

Cause

The NED mismatch error is detected due to the "Manufacture ID" value being recomputed (51496.SCN) with D/H02.15.20.00 and later code releases.  When doing an NDCL the new value ('GU' or 'BU') is returned on the RCD (Read Configuration Data) command.  The value on D/H02.14 and earlier code levels was incorrectly sending a value of '??'.  The abrupt changing of this value causes the host to declare a mismatch and will not bring the paths/CHPID (CHannel Path IDentifier) online.

Solution

The requirement to resolve the NED issue is to cause the host to forget the original NED (Node Element Descriptor) value.  This is accomplished by getting all paths/CHPIDs to the VSM from a given host offline concurrently.  As long as some path to the VSM is online to the host, the host will compare the NED value of the existing path to the NED value of any path trying to come online.  If the NED values do not match the new path will not be allowed to come online.

Provided below are two possible methods for minimizing the amount of time that all paths/CHPIDs to the VSM are required to be offline concurrently.

What could be done as an example (assumes all hosts have at least one path/CHPID in cluster 0 and another in cluster 1):

1. Perform the NDCL warm boot.  This upgrades the code in the VSM processors but not in the interface cards.
2. Take the paths/CHPIDs connected to cluster 0 (VCF/ICE00 to VCF/ICE03) offline.
3. GFR the cluster 0 interface cards to upgrade the interface code on those cards.
4. Take the paths/CHPIDs connected to cluster 1 (VCF/ICE10 to VCF/ICE13) offline.  It is at this point that all interfaces to the VSM are offline.  This allows the host to forget the NED value of the machine.
5. Put the paths/CHPIDs connected to cluster 0 (VCF/ICE00 to VCF/ICE03) online.
6. GFR the cluster 1 interface cards to upgrade the interface code on those cards.
7. Put the paths/CHPIDs connected to cluster 1 (VCF/ICE10 to VCF/ICE13) online.

Alternately, depending on host connectivity the following could be done.  In this example there are 4 hosts with each host having 2 paths/CHPIDs connected to the VSM and one of the 2 connections is in each cluster of the VSM.  This plan assumes the first host is connected to the VCF00 and VCF10 cards.  The second host would be connected to VCF01 and VCF11, etc., etc.

1. Perform the NDCL warm boot.  This upgrades the code in the VSM processors but not in the interface cards.
2. On the first host take the paths/CHPID connected to cluster 0, VCF/ICE00 offline.
3. GFR the cluster 0, VCF/ICE00 card to upgrade the interface code on that card.
4. Take the paths/CHPIDs connected to cluster 1, VCF/ICE10 offline.  It is at this point that all interfaces to the VSM from host 1 are offline.  This allows the host to forget the NED value of the machine.
5. Put the paths/CHPID connected to cluster 0, VCF/ICE00 online.
6. GFR the cluster 1, VCF/ICE10 card to upgrade the interface code on that card.
7. Put the paths/CHPID connected to cluster 1, VCF/ICE10 online.
8. Repeat steps 2 through 7 for the next host but note that the next host will use VCF/ICE01 and VCF/ICE11.  Repeat as necessary for remaining hosts.

This solution has no attachment