[apps::vmware::connector::plugin] --mode=vsan-cluster-usage: wrong unit for cluster.vsan.backend.latency

[joschi99]

Latest plugins. The backend latency unit returned from VSAN is not ms but µs. Please consider this for graphs (where unit will displayed with ms) and also for the thresholds.

[omercier]

Hi @joschi99 can you send an example of your command result in debug mode?

[joschi99]

Hi @omercier, here you can find a debug output

OK: Cluster 'XXXX' vsan backend read IOPS: 16599, write IOPS: 8232, congestions: 0, outstanding IO: 150.5 | 'cluster.vsan.backend.read.usage.iops'=16599iops;;;0; 'cluster.vsan.backend.write.usage.iops'=8232iops;;;0; 'cluster.vsan.backend.congestions.count'=0;;;0; 'cluster.vsan.backend.outstanding.io.count'=150.5;;;0; 'cluster.vsan.backend.throughput.read.bytespersecond'=641991025B/s;;;0; 'cluster.vsan.backend.throughput.write.bytespersecond'=159102785.5B/s;;;0; 'cluster.vsan.backend.latency.read.milliseconds'=4873ms;;;0; 'cluster.vsan.backend.latency.write.milliseconds'=287ms;;;0;
{"identity":"client-7067a6e354644fe7b207793b02417569","data":{"domain-c1053263":{"name":"XXXX","cluster_domcompmgr":{"latencyAvgRead":4873,"throughputRead":641991025,"iopsRead":16599,"iopsWrite":8232,"congestion":0,"oio":150.5,"latencyAvgWrite":287,"throughputWrite":159102785.5}}},"short_message":"OK","extra_message":"","vmware_connector_version":"3.2.5","code":0}
Cluster 'XXXX' vsan backend read IOPS: 16599, write IOPS: 8232, congestions: 0, outstanding IO: 150.5

As you can see the values for a returned but are not in ms but us

• "latencyAvgWrite":287
• "latencyAvgRead":4873 Real latency also from vCenter GUI are 0,2ms and 4,8ms

[joschi99]

any news on this?

[lucie-dubrunfaut]

Hello :)

I took the request into account and generated the associated ticket internally. We will discuss it with dev team soon.

[omercier]

Hi @joschi99, I'm working on this issue. I'll keep the value in its original unit, and display the same value as µs in both output and perfdata.

While rewriting the help section, I noticed that none of the filtering options below seem actually implemented.

=item B<--cluster-name>

cluster to check.
If not set, we check all clusters.

=item B<--filter>

Cluster name is a regexp.

=item B<--scope-datacenter>

Search in following datacenter(s) (can be a regexp).

I'm implementing the first one as a regular expression and remove the --filter that will become useless. Do you think the --scope-datacenter option has to be implemented or may it be just removed from the help?

Thanks in advance for your feedback!

[omercier]

Hi @joschi99, Can you also test this version of the plugin?

[joschi99]

Hi @omercier, done a test with following results:

OK: Cluster 'vsan-lab' vsan backend read IOPS: 19, write IOPS: 305, congestions: 0, outstanding IO: 7 | 'cluster.vsan.backend.read.usage.iops'=19iops;;;0; 'cluster.vsan.backend.write.usage.iops'=305iops;;;0; 'cluster.vsan.backend.congestions.count'=0;0:200;0:220;0; 'cluster.vsan.backend.outstanding.io.count'=7;0:500;0:2000;0; 'cluster.vsan.backend.throughput.read.bytespersecond'=231846B/s;;;0; 'cluster.vsan.backend.throughput.write.bytespersecond'=2811573B/s;;;0; 'cluster.vsan.backend.latency.read.milliseconds'=378µs;0:100000;0:250000;0; 'cluster.vsan.backend.latency.write.milliseconds'=275µs;0:100000;0:250000;0;
Cluster 'vsan-lab' vsan backend read IOPS: 19, write IOPS: 305, congestions: 0, outstanding IO: 7

1. The unit is now µs but the description is still milliseconds and not microseconds
2. We should have both filter possibilities to filter for datacenters and clusters, because on a vCenter you can have more then 1 datacenter and a single datacenter can contains more the 1 cluster

[omercier]

Hi @joschi99, Thanks for the feedback!

1. I'll fix this omission.
2. I think I was wrong about the filtering options, they are apparently passed to the connector itself and included in the query. I won't remove the options, then.

[omercier]

@joschi99 I have:

• fixed the metric name
• restored the --scope-datacenter option
• but I removed the --filter option, which is clearly not used in this mode. Any feedback still welcome :-)

[joschi99]

thanks, can you send me a limk to test the modified version?

[omercier]

The work in progress is in this branch.

[joschi99]

Have done a test, output seem's know ok and will be shown in microseconds: OK: Cluster 'vsan-lab' vsan backend read IOPS: 28, write IOPS: 344, congestions: 0, outstanding IO: 6 | 'cluster.vsan.backend.read.usage.iops'=28iops;;;0; 'cluster.vsan.backend.write.usage.iops'=344iops;;;0; 'cluster.vsan.backend.congestions.count'=0;;;0; 'cluster.vsan.backend.outstanding.io.count'=6;;;0; 'cluster.vsan.backend.throughput.read.bytespersecond'=344244B/s;;;0; 'cluster.vsan.backend.throughput.write.bytespersecond'=3499904B/s;;;0; 'cluster.vsan.backend.latency.read.microseconds'=401µs;;;0; 'cluster.vsan.backend.latency.write.microseconds'=312µs;0:1000;;0;

Filter for datacenter (--scope-datacenter) is working as expected (with regexp). Filter for cluster (--cluster-name) have some problem: when defined like --cluster-name='.*' (every cluster name/wildcard shoud be ok), then we will have back this error: UNKNOWN: Cannot find 'ComputeResource' object Same error will happen also when clustername is defined like --cluster-name='^vsan-lab$'