Skip to content
SHC Docs

shc status

show deployment and service health

Reference for the status CLI group. Source: modules/health/commands/.

SubcommandDescription
showshow health status for deployments and services

show health status for deployments and services

FlagTypeRequiredHelp
-a / --allboolnoshow all stacks; every tenant unless -t/--tenant names one
--all-tenantsboolnoshow every tenant’s deployments (platform admins; others see their own tenant)
--resourcesboolnoadd a table of live CPU and memory per stack

Scope, and how it matches shc ls:

Command lineDeployments returned
shc status -aevery tenant’s, with a TENANT column
shc status -a -t corptenant corp only
shc status -a with $SHC_TENANT=corp or shc ctx -t corptenant corp only
shc status -a --all-tenantsevery tenant’s, ignoring $SHC_TENANT / -t
shc status <stack>that stack in your own tenant

Reading across tenants requires the entitlement shc ls --all-tenants requires: the local UDS caller or a platform admin. Anyone else gets their own tenant, whatever they ask for.

Naming a stack is an ADDRESSING call and resolves in your own tenant, the way shc install and shc vault get do. Only the listing form widens.

--resources adds a second table holding what the running containers are actually consuming, sampled by the metrics tick:

STACK ENVIRONMENT CPU MEMORY MEASURED
mailcow prod 1.30% 273.0 MiB 14/14

This is a different number from shc quota status, which reports the deploy-time reservation the scope is charged. Measured on bitspur compose01 the same minute: shc quota status -t corp -s mailcow said 8.1 cores and 10.8 GB, while docker stats over the same 14 containers said 1.30% of one core and about 273 MiB. Both are correct; they answer different questions.

Read the MEASURED column. The figures come from the node-local metrics cache. Its peer fan-in merges every reachable node’s containers, so a follower is not structurally blind, but a peer the fan-in could not reach is skipped and its containers are simply absent. 2/14 means the row sums two of the stack’s fourteen services, which is real and is not that stack’s load. A stack nothing sampled reads unmeasured, never 0.00%: an idle container legitimately reads 0.00%, so that value cannot also mean “nobody looked”.

A daemon older than this change sends no figures at all, so every row reads unmeasured 0/<n>.

Source: modules/health/commands/