Skip to main content
← Back to list
01Issue
FeatureShippedSwamp CLIPublicTeam
Assigneesstack72

Relationships

↔ sibling #2465↔ sibling #3578

#4766 sub workflow placement on workers breaks some of my extensions my

Opened by svendowideit · 11/15/2026· Shipped 11/18/2026

@keeb asked for a user-facing highlevel version of #3578 and helped me write:

(this is an example, there are other cases i've hit in my extensions, and really, i want to know much much more)

what works

when using local swamp, i can use @SvenDowideit/caddy to install caddy, and have it call @SvenDowideit/sudo to setcap.

when using swamp serve (no workers) i can do the same.

what doesn't

if i try to run @SvenDowideit/caddy on a swamp worker from my swamp serve, then it fails.

because @SvenDowideit/sudo doesn't run on the worker

because i developed (like many new people) them on a local swamp, or on my non swamp worker swamp serve.

What I'd like

Certainty that swamp deployment choices don't change how a workflow works.

personally, i'd rather have to specify explicitly that sub-workflows can be placed on other nodes, showing that i've intentionally designed those for scalability, and for the default to be linear, as concurrency is hard.

02Bog Flow
✓OPEN✓TRIAGED✓IN PROGRESS✓SHIPPED+ 1 MOREASSIGNED+ 5 MOREREVIEW+ 12 MOREPR_LINKED+ 2 MORESESSION_SUMMARIZED

Shipped

11/18/2026, 1:06:58 PM

Click a lifecycle step above to view its details.

03Sludge Pulse
keeb moved this from a private collective to public11/18/2026, 2:20:19 AM
stack72 assigned stack7211/18/2026, 2:38:10 AM
stack72 linked sibling of #246511/18/2026, 1:02:59 PM
stack72 linked sibling of #357811/18/2026, 1:03:03 PM
Editable. Press Enter to edit.

system commented 11/15/2026, 5:23:46 AM

Classified automatically when this issue was filed.

  • Source: Swamp CLI

If you feel this classification is incorrect, add a ripple to tell us so.

stack72 commented 11/18/2026, 1:07:03 PM

Thanks @svendowideit for reporting this! We shipped: A nested workflow step passes its effective placement and worker pin to the child run by default, so a child runs where its caller runs on every deployment. The call step can opt out with placement own (today's behaviour) or place the child with its own target, labels or platform; a child that declares placement in its own YAML overrides what it inherited field by field and drops the inherited pin. Placement is resolved by one shared resolver at the call site, persisted on the child run for resume, and shown in run output.. The fix has been merged and a release is on its way. We appreciate your contribution to swamp.

Sign in to post a ripple.