This is a script to monitor the usage of a self-hosted GitHub runner by extracting the job information JSON from the diagnostics log located in the <runner-dir>/_diag directory.
Fields to extract are defined in the process.jq which caches the the extraced json. Before saving a basic filter is applied that converts the unreadable JSON from the serialization in the runner.
Formatting is applied on all cached JSON files at once using the filter in format.jq. All the magic happens in this file.
Relative values like the idle time are calculated, specified keys can be be removed or selected.
Output is grouped by date, table headers are inserted approximately for each page.
The table is displayed using column.
More output control is available via flags (see -h for defaults and more information):
| Flag | Name | Parameter | Description |
|---|---|---|---|
-k |
keep | field[,...] |
Display only selected fields |
-r |
remove | field[,...] |
Remove selected fields |
-s |
sort | field |
Sort by this field, @ to reverse |
-g |
group | field[=start[:end]][,...] |
Group by selected field at string slice |
-f |
filter | field=value[=mode][,...] |
Filter output by selected fields and values |
Filter modes
| Mode | Filter | Description |
|---|---|---|
| string | exact | Exact match |
| string | startswith | Match at start |
| string | endswith | Match at end |
| string | contains | Match if contains |
| number | less | Match if less |
| number | greater | Match if greater |
| date | after | Match dates after |
| date | before | Match dates before |
Prefix a mode with @ to invert the match.
If your script is in ~/actions-runner/diagnostics and your runner is in ~/actions-runner:
$ ./stats.sh -p ../_diag/Worker_2020I recommend to hardcode this path in stats.sh and set it as the default PREFIX value.
Then you can use the script like this:
$ ./stats.sh -r owner,event -f event=push,time=2020-01-01=after -g time=0:10,workflow -s @durationRemove owner and event fields, filter by event push and after the date 2020-01-01, group by the first 10 characters of time and then by the workflow name, descending sort by the duration.
Output:
2020-01-01
Workflow 1
time repo run actor build workflow job number queuing idle duration
2020-01-01T00:00:00Z repository-name 45674545454 myself 1 Workflow 1 Job 1 1 9 3600 120
2020-01-01T01:04:12Z repository-name 45674545456 myself 3 Workflow 1 Job 1 2 12 3600 110
Workflow 2
time repo run actor build workflow job number queuing idle duration
2020-01-01T00:02:00Z repository-name 45674545455 myself 2 Workflow 1 Job 1 1 129 0 120
2020-01-01T01:02:12Z repository-name 45674545457 myself 4 Workflow 1 Job 1 2 122 0 110
Values are just made up and not consistent.