Skip to content
Spinbox is experimental: APIs change and data may be lost. Status lists what is built and what proves it.

spin admin

View Markdown
Administrative commands for managing runners and registration tokens.
These commands require admin privileges.
Usage:
spin admin [flags]
spin admin [command]
Available Commands:
costs What the installation cost this month, by service, its forecast and its budget
health Whether anything is wrong with the installation, part by part
jobs The control plane's periodic work, and running a job now
logs Search what the installation's processes logged
runner Manage registered runners
settings The installation's settings: read them, and change one
shell A root shell on a runner, for a while: asked for, recorded, and the runner retired after
storage What the object store holds, and taking out what nothing needs
token Manage runner registration tokens
versions What is running where: this control plane, each runner and its machine
volume Re-key a workspace's disk, and list the disks hosts refuse
workspace Stop, recover or read the egress of a workspace that is somebody else's
Flags:
-h, --help[=true|false] help for admin
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin [command] --help" for more information about a command.
What the installation cost in a month, as its cloud's bill says, by service; what the month
is forecast to end at, and its budget. As the control plane last read the bill - a few times a day -
never the cloud itself.
Usage:
spin admin costs [flags]
spin admin costs [command]
Available Commands:
by What each workspace, team or owner cost this month, and what nobody did
Flags:
-h, --help[=true|false] help for costs
--month string the month, as 2026-10 (default: this month)
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin costs [command] --help" for more information about a command.
A month of the bill shared out: the runners by what each workspace held of them - cores and
memory, half each - and the volumes' bucket by what its volumes kept. What nobody caused is said
apart and never spread: the runners' capacity nobody held (idle), the bucket's bytes no workspace
keeps (images, unreferenced), and the platform's own cost by role and service. First, the standby:
what the installation costs a day with nobody working, estimated from the last billed day.
Usage:
spin admin costs by <workspace|team|owner> [flags]
Flags:
-h, --help[=true|false] help for by
--month string the month, as 2026-10 (default: this month)
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Each part of the installation as the control plane knows it - its database and the
database's archive, the log store, the collector, its jobs, its runners and the workspaces
waiting for one - with ok, degraded or failed, why, and as of when. A part the installation
does not have is not listed. Exits non-zero when any part has failed.
Usage:
spin admin health [flags]
Flags:
-h, --help[=true|false] help for health
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
The control plane's periodic work, and running a job now
Usage:
spin admin jobs [flags]
spin admin jobs [command]
Available Commands:
list Every job, how often it runs and how its last run went
run Run a job now rather than when it is next due
Flags:
-h, --help[=true|false] help for jobs
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin jobs [command] --help" for more information about a command.
Every job, how often it runs and how its last run went
Usage:
spin admin jobs list [flags]
Flags:
-h, --help[=true|false] help for list
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Run a job now rather than when it is next due. It is not waited for: spin admin jobs list
says how it went. Refused while it is running.
Usage:
spin admin jobs run <job> [flags]
Flags:
-h, --help[=true|false] help for run
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Search what the control plane, the proxy and the runners logged, newest first.
The query is Quickwit's query language; empty is everything in the window.
Examples:
spin admin logs
spin admin logs 'severity_text:error' --since 24h
spin admin logs 'service_name:runner AND body.message:"no host"' --limit 20
Usage:
spin admin logs [query] [flags]
Flags:
-h, --help[=true|false] help for logs
--limit int32 how many records, newest first: 1 to 1000 (default 100)
--since duration how far back to look, at most 744h (31 days) (default 1h0m0s)
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Manage registered runners
Usage:
spin admin runner [flags]
spin admin runner [command]
Available Commands:
activate Let a drained or cordoned runner take new workspaces again
drain Stop a runner taking new workspaces, until it is activated
list List all runners
policy What a runner may do, as against what it is
remove Take a runner of an autoscaling group out of the fleet: give it back to its group, or forget it
revoke Revoke a runner's access, for a machine that should never connect again
Flags:
-h, --help[=true|false] help for runner
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin runner [command] --help" for more information about a command.
Let a drained or cordoned runner take new workspaces again.
A revoked runner is not brought back by this: its machine cannot connect again.
Usage:
spin admin runner activate <runner-id> [flags]
Flags:
-h, --help[=true|false] help for activate
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Stop a runner taking new workspaces, until it is activated (spin admin runner activate).
Existing workspaces continue to run until stopped.
With --suspend, every running workspace is suspended off the runner: its memory
and disk go to the object store, and it is resumed on another runner on its own.
The command returns once each is suspended or has failed to be. With
--leave-stopped as well - a drain that shrinks the fleet - they stay stopped,
and each resumes on another runner when its owner starts it.
Usage:
spin admin runner drain <runner-id> [flags]
Flags:
-h, --help[=true|false] help for drain
--leave-stopped[=true|false] With --suspend: leave them stopped until their owners start them (a drain that shrinks the fleet)
--suspend[=true|false] Suspend the runner's running workspaces so they resume elsewhere
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
List all runners
Usage:
spin admin runner list [flags]
Flags:
-h, --help[=true|false] help for list
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Show or set what this installation decides about one runner.
A runner reads what it *is* - where the control plane is, who to trust, where it keeps its
data - from its own environment, before it can reach anything. Everything else it is told
here, on its heartbeat, and it reports back the revision it is running under: "declared" and
"running" being different for more than a moment is a runner that did not take it.
With no flags this shows the policy. Any flag writes it, and a write is the whole policy:
what you do not pass is the installation saying nothing about that setting, which leaves the
runner on its own default.
Usage:
spin admin runner policy <runner-id> [flags]
Flags:
--cpu-cores int32 cores this runner advertises; unset advertises what it has
--disk-gb int32 disk this runner advertises; unset advertises what it has
-h, --help[=true|false] help for policy
--max-workspace-cpu-cores int32 the cores one workspace may grow to there
--max-workspace-io-mb-per-second int32 MiB one workspace may read, and write, a second there; unset is no bound
--max-workspace-io-ops-per-second int32 I/O operations one workspace may read, and write, a second there; unset is no bound
--max-workspace-memory-mb int32 the memory one workspace may grow to there
--memory-mb int32 memory this runner advertises; unset advertises what it has
--preemptible[=true|false] a stop on this runner is the machine being taken away, not a restart
--preemption-metadata-url string where that cloud's metadata service answers; empty is its own link-local address
--preemption-source string the cloud that announces a reclaim: aws, gcp or azure (implies --preemptible)
--shutdown-grace duration how long this runner stays up after a stop signal so its workspaces can be moved off it
--writeback-background-bytes int dirty page cache at which this runner starts writing back (vm.dirty_background_bytes), below the limit
--writeback-limit-bytes int dirty page cache this runner holds before a writer waits (vm.dirty_bytes); unset is what it booted with
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Take a runner that is a machine of an autoscaling group out of the fleet.
A connected one is cordoned and given back to its group, which terminates the machine and is
asked for one less; it drains as it goes, and its row goes when it has. One that is not connected
is given back if its group still has its machine, and forgotten now, without waiting out
host_forget_after_hours.
A runner that is the only home of a workspace's disk is refused, naming them: stop them, which
publishes each disk, or move them first. A runner registered with a token is revoked instead
(spin admin runner revoke).
--taken is for a machine that may be compromised: its credentials are deleted and it is cut off
before its group is asked anything, and every team's disk whose key it was given is marked for
re-keying (spin admin volume rekeys lists them), as a revoke does. Its machine is given back all
the same. With --leave-workspaces it goes although workspaces' disks are only there: each runs
again only once recovered onto another runner (spin workspace recover).
Usage:
spin admin runner remove <runner-id> [flags]
Flags:
-h, --help[=true|false] help for remove
--leave-workspaces[=true|false] with --taken, remove although workspaces' disks are only on this runner; each runs again only once recovered elsewhere
--taken[=true|false] the machine may be compromised: revoke its credentials and mark every disk whose key it was given for re-keying
-y, --yes[=true|false] Skip confirmation
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Revoke a runner's access to the platform: for a machine being replaced, or one that may
be compromised.
The runner is disconnected and its credentials deleted: that machine cannot come back, and
the capacity returns only with a new machine. To stop a runner taking new workspaces for a
while, drain it instead (spin admin runner drain).
A runner that holds workspaces' disks is refused, naming them: each could then run only once
recovered onto another runner (spin workspace recover), losing what this one had not published.
--leave-workspaces revokes it anyway.
Usage:
spin admin runner revoke <runner-id> [flags]
Flags:
-h, --help[=true|false] help for revoke
--leave-workspaces[=true|false] revoke although workspaces are on this host; each runs again only once recovered elsewhere
-y, --yes[=true|false] Skip confirmation
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
The installation's settings: read them, and change one
Usage:
spin admin settings [flags]
spin admin settings [command]
Available Commands:
get Every setting as it is in force, or one
set Change one setting
Flags:
-h, --help[=true|false] help for settings
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin settings [command] --help" for more information about a command.
Every setting as it is in force - what the installation says, or its default - with where it is
changed and when a change reaches what it governs. With a key, that setting's value alone, as
'spin admin settings set' takes it.
Usage:
spin admin settings get [key] [flags]
Flags:
-h, --help[=true|false] help for get
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Change one setting, checked as the dashboard's save is: a number within its bounds, true or
false, one of its choices, a list as JSON (["a","b"]), a duration as 60s. The control plane
refuses anything else, and writes nothing.
Usage:
spin admin settings set <key> <value> [flags]
Flags:
-h, --help[=true|false] help for set
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
A root shell on a runner, for a while, with a reason.
The runner is emptied of its workspaces first - each suspended and resumed on another
runner - unless the shell is an incident's (--incident), which opens on the runner as it is.
Where the installation asks for it (host_shell_needs_approval), a second administrator
approves the shell before the runner is touched. Every session is recorded by the control
plane, kept in the installation's logs bucket under Object Lock. Once the shell has ended - its
minutes up, or closed - the runner is retired: revoked, every disk it held the key of marked
to be re-keyed, and its machine given back to its group, which replaces it.
Usage:
spin admin shell [flags]
spin admin shell [command]
Available Commands:
approve Approve a root shell another administrator asked for: its runner is emptied at once
close End a root shell: one not yet approved is refused; any other ends, and its runner is retired
list The installation's root shells on runners, newest first: every runner's, or one's
open Ask for a root shell on a runner, wait for it, and open a session on it
recording What a root shell's sessions recorded, asciicast v2, each session's in turn
Flags:
-h, --help[=true|false] help for shell
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin shell [command] --help" for more information about a command.
Approve a root shell another administrator asked for: its runner is emptied at once
Usage:
spin admin shell approve <shell-id> [flags]
Flags:
-h, --help[=true|false] help for approve
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
End a root shell: one not yet approved is refused; any other ends, and its runner is retired
Usage:
spin admin shell close <shell-id> [flags]
Flags:
-h, --help[=true|false] help for close
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
The installation's root shells on runners, newest first: every runner's, or one's
Usage:
spin admin shell list [runner-id] [flags]
Flags:
-h, --help[=true|false] help for list
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Ask for a root shell on a runner - or take the one asked for already - wait until it is open,
and open a session on it in this terminal. The session ends when the shell exits, this terminal
closes, or the shell's minutes are up. Opening another session on the same shell, while it is
open, is running this again.
Examples:
spin admin shell open 0199... --reason "the disk fills and nothing says why" --minutes 20
spin admin shell open 0199... --reason "it drops every stream" --incident
Usage:
spin admin shell open <runner-id> [flags]
Flags:
-h, --help[=true|false] help for open
--incident[=true|false] Open on the runner as it is, its workspaces running, without emptying it first
--minutes int32 How long the shell is open once it opens, up to the installation's host_shell_max_minutes (default 30)
--reason string Why: said to whoever approves it, and kept
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
What a root shell's sessions recorded, asciicast v2: each session's in turn, on standard output.
One session's plays with asciinema (asciinema play); the recordings stay in the logs bucket, under
host-shells/<shell-id>/, for as long as the installation keeps its logs.
Usage:
spin admin shell recording <shell-id> [flags]
Flags:
-h, --help[=true|false] help for recording
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
What the object store holds, and taking out what nothing needs
Usage:
spin admin storage [flags]
spin admin storage [command]
Available Commands:
explain What one key in the store is, and whose
list Every volume in the store, whose it is and whether it may be removed
remove Remove an image, an extension release, an unclaimed volume or an orphan from the store
untag-image Take a ref off the image it names
Flags:
-h, --help[=true|false] help for storage
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin storage [command] --help" for more information about a command.
What one key in the store is, and whose: a key copied off the bucket's console, or the
whole URI with the bucket before it. A key under a volume, or a layer a commit names, is
said with the volume it is of, as storage-gc last read the store.
Usage:
spin admin storage explain <key> [flags]
Flags:
-h, --help[=true|false] help for explain
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Every volume in the store, whose it is and whether it may be removed, as storage-gc last
read the store (it reads it daily; the dashboard reads it now).
Usage:
spin admin storage list [flags]
Flags:
-h, --help[=true|false] help for list
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Remove an image, an extension release, an unclaimed volume or an orphan from the store:
its row, then its volume, whose key is destroyed. Its layers go at storage-gc's next run.
An image its ref still names is untagged first (spin admin storage untag-image); the newest
release of an extension, one a workspace runs, and a workspace's disk are refused.
Usage:
spin admin storage remove <volume-id> [flags]
Flags:
-h, --help[=true|false] help for remove
-y, --yes[=true|false] Skip confirmation
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Take a ref off the image it names: no new workspace is given it by that ref, and the next
publish under the ref is what it names. The disks already cloned from it go on reading
it; once none does, it can be removed (spin admin storage remove).
Usage:
spin admin storage untag-image <ref> [flags]
Flags:
-h, --help[=true|false] help for untag-image
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Manage runner registration tokens
Usage:
spin admin token [flags]
spin admin token [command]
Available Commands:
create Create a runner registration token
list List registration tokens
revoke Revoke a registration token
Flags:
-h, --help[=true|false] help for token
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin token [command] --help" for more information about a command.
Create a token for registering a new runner: one-time, or with --reusable a pool's.
The token will be displayed only once. Save it securely.
A pool's token registers every machine that presents it until it expires or is revoked: what
an autoscaling group's machines take, with nobody at them. Keep it short-lived and rotate it
(the control plane's own machine can mint one with `controlplane registration-token`).
The policy flags are what every runner the token registers starts with, as `spin admin runner
policy` would set them: a spot group's runners need to know from their first heartbeat that
a stop is a reclaim. A runner registering again keeps what was decided about it since.
Examples:
spin admin token create --description "Host us-east-1"
spin admin token create --description "Team runner" --team engineering
spin admin token create --description "Temp runner" --expires 1h
spin admin token create --description "spot group" --reusable --expires 168h \
--preemption-source aws --shutdown-grace 100s
Usage:
spin admin token create [flags]
Flags:
--cpu-cores int32 cores each host it registers advertises; unset advertises what it has
--description string Description for the token
--disk-gb int32 disk each host it registers advertises; unset advertises what it has
--expires duration how long the token registers hosts; unset is the control plane's 24 hours, and at most 30 days
-h, --help[=true|false] help for create
--max-workspace-cpu-cores int32 the cores one workspace may grow to there
--max-workspace-io-mb-per-second int32 MiB one workspace may read, and write, a second there; unset is no bound
--max-workspace-io-ops-per-second int32 I/O operations one workspace may read, and write, a second there; unset is no bound
--max-workspace-memory-mb int32 the memory one workspace may grow to there
--memory-mb int32 memory each host it registers advertises; unset advertises what it has
--preemptible[=true|false] a stop on each host it registers is the machine being taken away, not a restart
--preemption-metadata-url string where that cloud's metadata service answers; empty is its own link-local address
--preemption-source string the cloud that announces a reclaim: aws, gcp or azure (implies --preemptible)
--reusable[=true|false] a pool's token: it registers every host that presents it until it expires
--shutdown-grace duration how long each host it registers stays up after a stop signal so its workspaces can be moved off it
--team string The team to scope the runner to, by id, path or me
--writeback-background-bytes int dirty page cache at which each host it registers starts writing back (vm.dirty_background_bytes), below the limit
--writeback-limit-bytes int dirty page cache each host it registers holds before a writer waits (vm.dirty_bytes); unset is what it booted with
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
List registration tokens
Usage:
spin admin token list [flags]
Flags:
-h, --help[=true|false] help for list
--include-used[=true|false] Include already-used tokens
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Revoke a registration token
Usage:
spin admin token revoke <token-id> [flags]
Flags:
-h, --help[=true|false] help for revoke
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
What is running where: this control plane, each runner and its machine
Usage:
spin admin versions [flags]
Flags:
-h, --help[=true|false] help for versions
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Re-key a workspace's disk, and list the disks hosts refuse
Usage:
spin admin volume [flags]
spin admin volume [command]
Available Commands:
refused The disks a host refuses to serve, and whose workspace each is
rekey Copy a workspace's disk under a key of its own and destroy the old one
rekeys The disks being re-keyed, and the ones a revoked host's keys marked for it
Flags:
-h, --help[=true|false] help for volume
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin volume [command] --help" for more information about a command.
List every disk its host refuses to serve, with why, the workspace that cannot run for it
and that workspace's owner. The listing is audited: it names workspaces that are not yours.
Usage:
spin admin volume refused [flags]
Flags:
-h, --help[=true|false] help for refused
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Copy a workspace's disk, whole, under a key of its own, point the workspace at the copy and
delete the disk - which destroys the key it was under, and every snapshot of the disk with it: those
are what that key opens. What to do when the key is assumed taken, by a host that held it.
The workspace is anybody's, named by its id. It must be stopped or quarantined. A disk something
was cloned from is refused.
Usage:
spin admin volume rekey <workspace-id> [flags]
Flags:
-h, --help[=true|false] help for rekey
-y, --yes[=true|false] Skip confirmation
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
The disks being re-keyed, and the ones a revoked host's keys marked for it
Usage:
spin admin volume rekeys [flags]
Flags:
-h, --help[=true|false] help for rekeys
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Stop, recover or read the egress of a workspace that is somebody else's
Usage:
spin admin workspace [flags]
spin admin workspace [command]
Available Commands:
egress What somebody else's workspace has reached beyond this installation
recover Move somebody else's workspace off a host that is gone
stop Stop somebody else's workspace
Flags:
-h, --help[=true|false] help for workspace
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Use "spin admin workspace [command] --help" for more information about a command.
List every destination a running workspace that is not yours has been allowed to reach,
as its owner's 'spin workspace egress' does: the host's counters by destination, and nothing of
what was carried. The reading is audited, with the reason.
Usage:
spin admin workspace egress <workspace-id> [flags]
Flags:
-h, --help[=true|false] help for egress
--reason string Why: kept on the audit entry for the reading
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Move a workspace that is not yours to another host after the one it was on died, as its
owner's 'spin workspace recover' does.
The cost is the work the dead host had written and not yet published, and it is the owner's
work. Run without --accept-data-loss to see that cost; nothing is changed until you pass it.
--reason is required, and the server refuses a recover without one: it is kept on the audit
entry, and the owner reads it, with who asked, on the workspace's history.
Usage:
spin admin workspace recover <workspace-id> [flags]
Flags:
--accept-data-loss[=true|false] go ahead and give up, for its owner, the work the old host had not published
-h, --help[=true|false] help for recover
--reason string Why (required): kept on the audit entry, and on the workspace's history for its owner
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false
Stop a workspace that is not yours, as its owner's 'spin workspace stop' does: what to do
with one whose stop could not publish its disk, which runs on and holds its host.
--force gives up what only its host has rather than publish it first. The stop is audited with
the reason, and the workspace's own history names you as who stopped it.
Usage:
spin admin workspace stop <workspace-id> [flags]
Flags:
--force[=true|false] give up what only its host has rather than publish it first
-h, --help[=true|false] help for stop
--reason string Why: kept on the audit entry for the stop
-y, --yes[=true|false] Skip confirmation
Global Flags:
--config string config file (default is $HOME/.spin/config.yaml)
-o, --output string text for a person, or json: one document of what the API answered, and no prose (default "text")
--profile string named profile to use; SPIN_PROFILE names it when this flag does not (default "default")
--traceparent string the W3C traceparent of the span this command is part of; its own span is sent to the installation unless the profile says telemetry: false

Generated by task reference from spin/cmd/cli at 51f7e0dea.

Installations run Spinbox v20261027.01; the reference is generated from spin v20261114.01 (51f7e0dea). This page as Markdown, for a model or a terminal.