Missing some required metrics and bug collect metrics

[an16112000]

I have configured the zHMC exporter to collect metrics according to the vendor's documentation. However, the returned metrics are currently missing two pieces of information that I need:
+Differentiation between ROCE and OSA cards
+Which FC card is each LPAR using? Which port? And what is the utilization?
+The zHMC exporter takes a long time to collect metrics related to adapters (about 8 minutes between the two most recent collections), even though all metrics have been configured with an interval of 30 seconds.
The zHMC is a z16 running version 2.15.

[andy-maier]

@an16112000 Hi. Can you please provide some more information:

• Is your CPC in classic mode or DPM mode?
• For differentiation between ROCE and OSA cards, which metric or metric group of the exporter is this for?

On "Which FC card is each LPAR using? Which port? And what is the utilization?":

If this is a CPC in DPM mode, the exporter metric group "adapter-usage" provides the utilization in metric "zhmc_adapter_usage_ratio". For CPCs in classic mode, the exporter metric group "channel-usage" has metric "zhmc_channel_usage_ratio". However, in both cases this is at a level of the entire adapter, and not at the level of what a particular LPAR uses. Note that for FICON Express adapters, each port is represented as an adapter object / channel, so each of the metric values is actually for an adapter port.

If you want the utilization of a FICON port caused by each LPAR, a new metric would need to be implemented in the firmware. This is nothing the exporter project can provide for you, unless it is implemented in the firmware. Do you have a contact into IBM Z where you can open such a requirement?

[andy-maier]

On the issue with taking a long time to collect metrics related to adapters:
Which version of the exporter are you using?

Starting with version 2.2.0, the resource caching was improved. If you are not yet on version 2.2.0 or higher, please use that and see if the problem is reproducable.

If you are already at version 2.2.0 or higher, then please provide a log of the exporter. Use its command line options --log file.log --log-comp all=debug to create a log file, and perform at least two collections that take a long time. The log file does not contain any sensitive information, so you can attach it to this issue. If you don't want to do that, let me know and we find some solution for transferring the file.

[an16112000]

So, how about separate between ROCE and OSA of adapter in DPM mode? and how its utilization?

[andy-maier]

On ROCE and OSA cards on machines in DPM mode:

• For OSA cards, the HMC supports metrics at the NIC level (exported as "zhmc_nic_...", exporter metric group "partition-attached-network-interface) and at the adapter port level (exported as "zhmc_port_...", exporter metric group "network-physical-adapter-port"). The adapter name and port index are available on each metric as labels, so you can differentiate the different adapter cards and ports.
• For ROCE cards, the NIC level and port-level metrics are not supported by the HMC. If you need that, you would need to open a PMR with IBM.

For both OSA and ROCE, the adapter utilizations are available in metric "zhmc_adapter_usage_ratio" (exporter metric group "adapter-usage"). That metric has a label for the adapter name.

Let me know in case that does not answer your questions. Do you want an indicator for the adapter family as opposed to adatper names?

[andy-maier]

@an16112000 I have to correct myself: The port and NIC metrics for RoCE adapters were missing due to a bug which has been fixed in z16 bundle S46. Using that bundle (or higher) should cause the HMC to provide port and NIC metrics for RoCE adapters, and the exporter should automatically surface these metrics without any further change to it.

[andy-maier]

@an16112000 Please let me know whether the answers above addressed your issue. We'll keep this issue open for a couple more weeks.