Skip to content

Sinks & Placement

A Sink is a named execution target: a delivery definition you create — a name plus a placement — or an unnamed machine registration (the row a node created when its agent registered). The Sinks screen lists your tenant’s sinks plus the platform’s pool rows, grouped as platform. Pool by default, pin by choice — an unpinned sink balances across your tenant’s live sink nodes; pinning restricts it to the one node.


On the Sinks screen (or inline in the flow builder’s “New sink…” slide-over):

  • Name it — registrations are unnamed until you name them; give it a name that says what it does.
  • Pool (default) — leave it unpinned and it balances across the tenant’s live sink nodes. Add attribute constraints (e.g. zone=ot) to narrow the pool to nodes carrying those tags.
  • Pinned to a node — pick one of your sink nodes. Steps on this sink always run on that node.

Removing a sink offers three paths, matching the Sinks screen: Disable (the row goes OFFLINE — out of dispatch, kept for event history; its flows stay attached and stop delivering until it is re-enabled), Repoint & delete (every referencing flow moves to its tenant’s platform row, then the row is deleted), and direct delete for a sink no flow uses.


For every step, the hub checks one thing: the step’s constraints are a subset of a node’s tags (its reported os/arch/version/engines/libraries plus user-assigned attributes).

  • The step’s needs fall out of the action’s library requirements (the platform library declares them: engine, minimum version, OS families — see Actions & Libraries).
  • The sink’s constraints add attribute filters.
  • A pool sink picks one node from the matching pool, load-balancing across it, and fails over inside the pool when one goes offline.
  • A pinned node never falls back to the platform; work with no match is held and flagged on the flow.

Full details: preset catalog.


Your automations run on nodes (see Nodes). When you create a node, the platform generates a Client ID and Client Secret — the secret is displayed only once (regenerable, invalidating the old one). These are what you provide to the installed service:

CredentialEnvironment VariablePurpose
Client IDCLIENT_IDIdentifies which node the agent belongs to
Client SecretCLIENT_SECRETAuthenticates the agent to the Hub

Install a sink agent with the Installing Sinks guides.


Nodes carry an attribute sheet: reported attributes (os, arch, engine version, libraries — written by the node at registration, read-only) and user-assigned attributes (e.g. zone, tools) with value lists you manage on the Nodes screen. Values are sanitized and deduplicated; renaming a value updates every node and constraint at once. See Core Concepts.