Connectors overview
A connector is a saved connection to a database, warehouse, or endpoint — the thing your queries actually run against. This page covers the parts common to every connector; for the exact fields a specific connector asks for, see its own page under Connecting Data.
Creating a connection
Add a connector, pick its type, and fill in its connection fields (host, credentials, database name — the exact set varies by type). Passwords and other secrets are stored encrypted and are never echoed back by the API once saved — the edit form shows every other field, but not the secret itself.
[!NOTE] Creating a connector requires having at least one group that both grants the create-data-source permission and is itself unrestricted. If your access is limited to specific connectors, ask a workspace admin to widen it rather than trying to create one yourself.
Testing a connection
Before or after saving, you can test a connection — arcodash opens it and runs a trivial query
(effectively SELECT 1, or the connector’s closest equivalent) to confirm the credentials and
network path work. A test can run against a not-yet-saved configuration (useful while filling out
the form) or against an already-saved connector’s stored options.
A failed test returns a human-readable error — bad credentials, an unreachable host, or a misconfigured field — without ever leaking the configured secrets back into the error message.
Connection health
Each connector shows a health indicator next to its name, with one of four states:
| State | Meaning |
|---|---|
| Off | Turned off — see Turning a connector on and off below. |
| On · Healthy | The most recent test or query against this connector succeeded. |
| On · Not checked yet | Saved, but nothing has connected to it yet — no test run, no query executed. |
| On · Unreachable | The most recent test or query against this connector failed. |
A successful Test connection — whether run while creating the connector or later from its detail page — is what actually sets Healthy; simply saving the form does not. A brand-new connector starts at “Not checked yet” until either a test is run against it or a query executes successfully.
Schema introspection
Once connected, arcodash can introspect the schema: every table/collection/index visible to the configured user, with its column or field names. This populates the schema browser in the query editor so you can see what’s queryable without leaving arcodash. Schema is cached rather than fetched on every page load — request a refresh explicitly when the underlying connector’s structure has changed, rather than waiting for it to happen automatically.
[!NOTE] If nothing’s cached yet, opening the schema browser triggers an automatic first fetch — but only for someone whose role can edit the connector. A viewer without that permission just sees an empty schema in the same situation, until a teammate who can edit the connector opens it and the fetch happens on their behalf; from then on, everyone sees the same cached schema. This is why a colleague can see a table you can’t, the first time either of you opens a freshly connected connector.
Turning a connector on and off
A connector can be turned off, then turned back on later, without losing its configuration — turning it back on picks up right where it left off, no re-entering credentials. This shows as an Off state on the connector’s health indicator (see Connection health above) rather than a delete.
[!TIP] Turn off a connector you know is temporarily unreachable (a maintenance window, a credential rotation in progress) instead of leaving it to fail health checks and confuse anyone else looking at it.
Under the hood this is called pausing — the API’s paused and pause_reason fields are what
Off reflects — but the button you’ll actually click is labeled Turn off (and Turn on
to reverse it).
Editing vs. deleting
Editing a connector updates its name, type, or connection options in place — anything built on top of it (queries, dashboards) keeps working against the same connector id. Deleting removes the connector entirely; there’s no undo, so turning it off is usually the safer move if you’re not sure you want it gone for good.
Access is scoped by group
Not everyone in an organization necessarily sees every connector — which ones a user can see and query is determined by their group memberships, not set per-user. See Groups and connector access for how group-based connector access control works, including the restricted/unrestricted distinction and how access from multiple groups combines.
Where to go next
- Query Results (meta-source) — join across your other queries’ cached results without a real database connection.
- Data observability monitors — watch a connector’s tables for freshness, row-count, null-rate, and duplicate problems; the connector’s detail page can suggest these checks from its schema with AI.
- Data handling & privacy — exactly what arcodash sees, caches, and encrypts once a connector is saved.
- Pick a connector from the sidebar for its exact connection fields and example configuration.