stratumctl admin reconcilers
Inspect the reconciler fleet (bootstrap-admin only)
stratumctl admin reconcilers [flags]Aliases: reconciler
Subcommands
Section titled “Subcommands”stratumctl admin reconcilers drift: Show which items diverged and why each is not convergingstratumctl admin reconcilers fleet: Cluster-wide reconciler status (which node runs each one)stratumctl admin reconcilers list: List every reconciler with health, drift, lock state and observed cadence
stratumctl admin reconcilers drift
Section titled “stratumctl admin reconcilers drift”Show which items diverged and why each is not converging
stratumctl admin reconcilers drift <name> [flags]Dump one reconciler’s per-item drift.
“drift: 3” says three things diverged. This says which three, and why each one is not converging: awaiting_confirmation (with the streak, so a blip reads differently from something wedged for an hour), backoff (with the apply error that caused it), circuit_breaker, dry_run, a suppressed reap, apply_failed, blocked_on_owner (waiting on the tenant, e.g. a bucket delete parked because the bucket still holds objects; counted as blocked, not drift, and never degrades health), or acting.
COVERAGE is the other half, and the part a drift diff cannot supply. It reports the desired-state status histogram and whether the reconciler’s work query actually selects each status. A non-terminal status the query does not select, holding rows that keep aging, means those rows are reconciled by nobody — the shape, where delete_blocked buckets were stranded forever while drift read 0.
Examples
Section titled “Examples”stratumctl admin reconcilers drift dbstratumctl admin reconcilers drift db --output json| Flag | Type | Default | Description |
|---|---|---|---|
-o, --output |
string | text |
output format: text | json |
stratumctl admin reconcilers fleet
Section titled “stratumctl admin reconcilers fleet”Cluster-wide reconciler status (which node runs each one)
stratumctl admin reconcilers fleet [flags]Show every reconciler as the CLUSTER sees it.
Prefer this over list for “is this reconciler working?”.
list reports the in-memory state of whichever control plane answered and
does not say which node that was, so on a multi-node cluster it gives a
different answer depending on where the request lands. In particular a
leader-gated reconciler reads never_ran on every node except the one holding
its lock — correct, and unreadable without knowing which node is leading.
This view names it.
Health rolls up differently by shape, because the right answer differs. A leader-gated reconciler takes the LEADER’s health: a follower skipped the pass, so its never_ran says nothing about the work. A per-node reconciler takes the WORST across nodes, since one broken node is genuinely broken.
STALE is a verdict, not a gap. A node that stops reporting is excluded from the rollup rather than contributing its last-known-good health, and if nobody has reported recently the answer is unknown — never converged. That is what stops “one node has been down an hour” hiding behind “converged everywhere”.
Examples
Section titled “Examples”stratumctl admin reconcilers fleetstratumctl admin reconcilers fleet --output json| Flag | Type | Default | Description |
|---|---|---|---|
-o, --output |
string | text |
output format: text | json |
stratumctl admin reconcilers list
Section titled “stratumctl admin reconcilers list”List every reconciler with health, drift, lock state and observed cadence
stratumctl admin reconcilers list [flags]List the reconciler fleet as this control plane sees it.
Reconciler status is PER NODE. A reconciler shown as skipped here may simply be leader-gated — another node holds its advisory lock and is doing the work. Query all nodes before concluding anything; a symptom on one node is not a fault.
The CADENCE column is the one that catches a trigger storm: it compares the observed gap between the last two ticks against the configured interval. A reconciler firing far above its configured rate is as broken as one not firing at all, and nothing surfaced it before.
STALE marks a health that describes an older pass — the reconciler has ticked since, but those ticks did no work. never_ran means no pass has ever executed on this node, so there is nothing to report; it is not a claim that anything is wrong, and it is not a claim that anything is right.
Examples
Section titled “Examples”stratumctl admin reconcilers liststratumctl admin reconcilers list --output json| Flag | Type | Default | Description |
|---|---|---|---|
-o, --output |
string | text |
output format: text | json |