Skip to main content
Preview. This part of the console is still being built. What you see here describes the intended behavior and can change before general availability.

Outpost FAQ

An evaluation of an outpost tends to stall on the same handful of questions, usually asked by someone who was not in the first meeting. Here are the short answers, with links to the pages that explain them.

Platforms​

Can it run on AWS Fargate?​

No. The outpost runs every scan as a set of containers on the host's Docker Engine, through the Docker socket, and a Fargate task has no Docker to drive and no disk that outlives it. That is why the AWS deployment uses ECS with EC2 capacity: one instance whose Docker the outpost drives. See Deploy on AWS ECS.

Can it run on Kubernetes?​

Not as a Kubernetes workload today. The outpost has no way to start a scan's containers through Kubernetes, only through Docker Engine. Run it on a Linux virtual machine or instance with Docker beside your cluster. It reaches the same registries and Git servers your cluster does, as long as it sits in the same network.

Can it run on Windows or macOS?​

No. The outpost is a Linux program and needs Docker Engine on Linux, on arm64 or x86_64.

Does it run on arm64?​

Yes. The outpost image holds both arm64 and x86_64, and the AWS ECS stack defaults to arm64.

Networks​

Does it need a public IP address or an inbound rule?​

No. It only makes outbound connections. On AWS, a private subnet behind a NAT gateway or a proxy is the usual place for it.

Can it run in an air gapped network?​

Not in one with no route out at all. The outpost needs outbound HTTPS to your Dwaar console for its work, and to Dwaar's registry and results storage for the analyzer images and the results. An HTTPS proxy is enough: everything the outpost and its analyzers reach goes through it. See Outbound connections to allow.

Does it work behind a TLS inspecting proxy?​

Yes, on a Linux host: enroll it with --ca-file naming your proxy's CA. The AWS ECS stack cannot take a CA yet. See Proxies and TLS inspection.

Can Dwaar reach into my network through it?​

No. Dwaar only answers the outpost's own requests for work. Nothing connects to the outpost, and it carries no traffic from Dwaar to your systems.

Tenants, teams and connections​

Can one outpost serve several Dwaar tenants?​

No. An outpost enrolls into one tenant with an enrollment token from that tenant, and takes work only from it. A service provider running Dwaar for several customers needs an outpost per customer tenant, and should give each its own host.

Can one outpost serve several organizations in my tenant?​

Yes. An outpost belongs to the tenant, and connections from any organization can bind to it. Its pane lists the connections bound to it in your organization scope.

Can one connection use two outposts, for load or for failover?​

No. A connection runs on exactly one outpost. Spread load by binding different connections to different outposts, and recover from a lost outpost by moving its connections. See Availability.

Which connections cannot run on an outpost?​

GitHub App connections, because their tokens are signed with Dwaar's app key; use a token connection instead. And a secrets manager only the outpost can reach, because Dwaar reads secrets managers itself. See What does not run on an outpost yet.

What happens to findings when I move or unbind a connection?​

Nothing. Findings belong to resources, not to where they were scanned. Past runs keep the place they ran in the queue's history, and new work goes to the new place.

Operations​

Does it update itself?​

No. You upgrade it when you choose, and Dwaar accepts an outpost one protocol version behind its own, so you have time. See Upgrading.

Can Dwaar's staff log in to it?​

No. Nothing connects to it, and Dwaar holds no credential for the host.

How large should the host be?​

Large enough for the peak memory of the scans you send, by their limits: about 30 GiB for a repository scan with every category on, 12.5 GiB for a container image scan of one platform. Plan an outpost shows the arithmetic and three worked sizes.

What happens when the outpost is down?​

Discovery and tests of its connections fail at once and run again at the next discovery. Scans wait for it, for up to about an hour, then fail. Nothing moves to another outpost by itself. See How an outpost works.

Cost​

What does an outpost cost to run?​

The host is yours, so you pay for it as for any other: the machine, its disk, and the outbound traffic, starting with about 16 GB of analyzer images on the first pull. On AWS, the stack's cost is the instance and its disk, one Secrets Manager secret and the logs; the instance, running all the time, is most of it. See Cost.

Is a smaller host cheaper in practice?​

Only if it accepts the scans you send. A host too small for a repository scan with every category on refuses it with InsufficientCapacity, and no amount of waiting helps. Size for the scans first, then for the price.

Try this next​

Take the question your evaluation is stuck on, find its answer here, and follow the link to the page behind it. If the question is not here, it is worth asking Dwaar support: the next reader probably has it too.