metrics

read onlyrepeatable

What applications are using right now: processor, memory, and how many instances are reporting.

Every figure is per instance, because the limit is per instance too. An application running five copies at half its memory is using five halves of one limit, not two and a half limits.

Usage

outplane metrics [APP]

Arguments

Argument	Type	Description
`APP`	`string`	Application name or id. Omit for the whole team.

Output

Field	Type	Description
`app`	`string`
`cpuMillicores`	`int`	1000 millicores is one core.
`cpuPercent`	`int \| null`	Against the per-instance limit. null when no limit is on record, and above 100 when the app is over it.
`memoryBytes`	`int`	Working set, per instance.
`memoryPercent`	`int \| null`	Against the per-instance limit.
`networkInBps`	`float`	Bytes per second received.
`networkOutBps`	`float`	Bytes per second sent.
`diskReadBps`	`float`
`diskWriteBps`	`float`
`instances`	`int`	How many instances are reporting, which is what is running. It differs from the configured count while an app is starting, stopping or partly crashed.
`cpuLimitMillicores`	`int`	What one instance may use.
`memoryLimitMb`	`int`	What one instance may use.

Examples

What everything is using

outplane metrics

One application

outplane metrics checkout

Find what is near its limit

outplane metrics --json --fields app,cpuPercent,memoryPercent

{
  "items": [
    {
      "app": "checkout",
      "cpuPercent": 12,
      "memoryPercent": 83
    },
    {
      "app": "worker",
      "cpuPercent": 0,
      "memoryPercent": 4
    }
  ],
  "total": 2,
  "truncated": false
}

What to Know

Every figure is per instance, not per application: usage is averaged across the instances, because the limit is per instance too. An app running five replicas at 50% is using five halves of a limit, not 250% of one.
instances is what is reporting, which is not the configured replica count. app list has the configured one, and a difference between them means the app is starting, stopping or partly crashed.
cpuPercent and memoryPercent are null when no limit is on record, and never capped: over 100 is a real state and the most useful thing this reports.
An application with no running instances reports zeroes rather than being omitted, so a paused app is visible as a paused app.
Naming one application filters the rows, not the queries. Cost is the same whether you ask about one or all of them, so ask once rather than in a loop.
Rates are two-minute averages. Two calls a second apart return nearly the same numbers, and neither shows a spike that lasted seconds.
The text table writes bytes and rates in units a person reads. The machine formats carry the raw numbers, so nothing has to be parsed back.

Errors

Beyond 0 for success, this command exits with:

Exit	Kind	Meaning
`2`	`usage`	Invalid arguments, unknown flag, or client-side validation failure.
`3`	`auth`	Not authenticated, token revoked or expired, or forbidden for this team.
`5`	`not_found`	The named resource does not exist, or is not visible to this credential.
`8`	`upstream`	The Out Plane API returned a server error.

The code on the error object is one of logs.no_team_slug, app.not_found, usage.empty_argument, metrics.bad_response, metrics.query_refused, context.no_team. Branch on that, or on the exit status. The message is prose and changes.