Skip to:
When using a filter on service name with --instance paramter, every metrics such as “node_*” will be excluded.
--instance description is "Service name to filter instances. Use multiple times to filter by multiple instances"
--instance
https://github.com/percona/pmm-dump/blob/e85f85c441cefaccaa9048dd88c403d4a4e2373d/cmd/pmm-dump/main.go#L82
It works as expected, as it will add the correct filter to the selectors:
https://github.com/percona/pmm-dump/blob/e85f85c441cefaccaa9048dd88c403d4a4e2373d/cmd/pmm-dump/main.go#L190-L192
Problem is:
Any node_* metrics (so every metrics from OS dashboards) do not have service_name label, so they won’t match the query
For example, labels for node_cpu_seconds_total
node_cpu_seconds_total
A quick test on explore confirms it, filtering with a service_name that does exist won’t work:
The issue is reproducible using pmm-dump GUI from PMM(tested with version 2.41.2)
Using the pmm-dump(0.7.0) tool instead, the issue did not happen.
Attached are both dumps from personal tests.
When using a filter on service name with --instance paramter, every metrics such as “node_*” will be excluded.
--instancedescription is "Service name to filter instances. Use multiple times to filter by multiple instances"https://github.com/percona/pmm-dump/blob/e85f85c441cefaccaa9048dd88c403d4a4e2373d/cmd/pmm-dump/main.go#L82
It works as expected, as it will add the correct filter to the selectors:
https://github.com/percona/pmm-dump/blob/e85f85c441cefaccaa9048dd88c403d4a4e2373d/cmd/pmm-dump/main.go#L190-L192
Problem is:
Any node_* metrics (so every metrics from OS dashboards) do not have service_name label, so they won’t match the query
For example, labels for
node_cpu_seconds_totalA quick test on explore confirms it, filtering with a service_name that does exist won’t work: