StorageTek Virtual Storage Manager System 6 (VSM6)
 
StorageTek Virtual Storage Manager System 7 (VSM7)
 

PLA-Support>Sun Systems>TAPE>Virtual Tape>SN-TP: VSM6
 

In this Document

Symptoms

Changes

Cause

Solution

Created from <SR 3-12897757981>

Applies to:

Symptoms

Symptoms may include any of the following:

CU RESOURCES EXCEEDED – INIT FAILURE – NO RESOURCES AVAILABLE

IOS125I messages similar to the following example:
IOS125I CHPID 38 IS EXPERIENCING MULTIPLE I/O TIMEOUTS CAUSING PROLONGED INITIALIZATION.

Changes

 A new HCD/IOCP configuration has been implemented for the purpose of adding additional LPARS, CHPIDs or other hardware.

Cause

The VSM6 or VSM7 has been over-subscribed with respect to logical paths to a specific FICON port within the product.
This means that an attempt to create more than 1024 logical paths was made.
 

Solution

Logical paths within the VSM6 and VSM7 products is a reference to a logical path created between the machine and the host/LPAR.
These paths exist at the rate of 1 per VCU (Virtual Control Unit) and not one per device.
The VSM6 and VSM7 both support up to 16 VCUs which is used in nearly all cases.
There is a limit of 1024 logical paths per FICON port in these products.

The calculation for how many logical paths are in use on a given port is:

16 VCUs (the default) X the number of LPARs configured on the CHPID going to that port = the number of logical paths for that port.

Consequently, assuming 16 VCUs per LPAR, each port can support up to 64 LPARs.
(16 X 64 = 1024)

NOTES:
1. It is the number of LPARs and not the number of physical hosts that must be used for this calculation.
2. Only one CHPID connection per LPAR per FICON port is supported in the VSM6 or VSM7.

The solution to having too many logical paths (greater than 1024) configured to a single port of the VSM6 or VSM7 is to correct the physical configuration and/or the HCD/IOCP and then re-IPL.

Logical Path Oversubscription Diagnosis:

It is possible for a customer environment to attempt to establish more than 1024 logical paths per port. For example a large site might mistakenly configure a second CHPID to a single port for more than 32 LPARs which would effectively double the number of logical paths that port would attempt to create and this would exceed the 1024 logical paths limit.

Within the tikka.out FICON port logs there are 3 entry types of interest relative to this subject.

1. ELP Received (Establish Logical Path) - Example:
2016-06-20 06:39:46.569159 p0>SBCU - ELP Received by SB(lp added) - ch_img:3, cu_img:5, lp:10000003FF, loop_id:3 on Port 0.

This indicates a logical path has been established.

2. RLP (Remove Logical Path) - Example:
2016-06-19 03:26:59.301085 p0>SBCU - RLP - ch_img:5, cu_img:F, lp:6000000AF, loop_id:3 on Port 0.

This indicates a logical path has been removed (which makes that path number available to be established again).

3. ELP Failure - Example:
2016-06-19 00:59:18.640092 p0>SBCU - ELP Failure Out Of Resources: CH 0x03, CU 0x0A, N_Port Id 0x010200, loop id 2

This means that all available logical paths have already been established.

Decoding - With respect to the above examples:

cu_img = the VCU number

lp:*000000*** = the logical path number in hexadecimal (where * = some hexadecimal number (only the last 3 characters count). The first logical path to be established will be path number zero. Consequently the highest path number that should ever be found is x3FF (x3FF = 1023 decimal).

A simple grep of the tikka.out files for ‘ELP Failure’ can verify whether or not there has been an attempt to establish more than 1024 logical paths.

The tikka.out files are available in a set of statesave (collect) files. Alternately, if direct access to the machine is available then the same grep can be performed live. The tikka.out files are in the /var/vsm/output/current directory.

Diagnosing on the machine:
1. Connect to the node in question via putty.
2. Change to the correct directory:
cd /var/vsm/output/current
3. Grep for the ELP Failure messages:
grep ‘ELP Failure’ tikka.out.p*
4. If the grep finds ‘ELP Failure’ messages then it verifies that an attempt (or multiple attempts) were made to establish too many (greater than 1024) logical paths.

5. Additional grep commands can be used to further isolate the problem to a specific port if needed.
Example:

echo -n Port 0 ELP Failure count: ; grep 'ELP Failure' tikka.out.p0* | wc -l ; echo -n Port 1 ELP Failure count: ; grep 'ELP Failure' tikka.out.p1* | wc -l ; echo -n Port 2 ELP Failure count: ; grep 'ELP Failure' tikka.out.p2* | wc -l ; echo -n Port 3 ELP Failure count: ; grep 'ELP Failure' tikka.out.p3* | wc -l

The command string above would return something like this (if we only had the problem on two ports):
Port 0 ELP Failure count: 2414
Port 1 ELP Failure count: 0
Port 2 ELP Failure count: 1371
Port 3 ELP Failure count: 0

This solution has no attachment