Spaces

Fixes for spaces that will not launch, start, resize or show up in your list.

Launch is disabled on a personal manifest

You see: Only the owner can launch their personal manifest

Cause: A personal manifest launches only for the person who owns it.

Fix: Ask the owner to send you a copy. In the Manifests page, the owner opens the row menu and chooses Send a copy. The copy has no credentials attached, so you add your own. The owner can also move the manifest to a team or to the organization.

Launch is disabled without permission

You see: You don't have permission to launch this manifest

Cause: The manifest belongs to a team or the organization, and you do not have use access to it.

Fix: Ask a lead of the owning team, or an org admin, to share the manifest with you as Use. See Sharing: See and Use.

Launch is refused: manifest not usable

You see: you can see this manifest but you are not allowed to launch it

Cause: The manifest is shared with you as See. Launching needs Use.

Fix: Ask a lead of the owning team, or an org admin, to change your share to Use.

Launch is refused: manifest not found

You see: manifest not found

Cause: The manifest does not exist, or it is not shared with you.

Fix: Run marshal manifests list or open the Manifests page to find a manifest you can see. If you expected this one, ask its owner to share it with you.

Launch is refused: the stack was unpublished

You see: A message that ends this manifest's stack has been unpublished by an admin

Cause: An admin unpublished the stack behind the manifest. The manifest itself is kept.

Fix: An admin republishes the stack in the Stacks page, or you point the manifest at another published stack. Then launch again.

Launch is disabled until credentials are added

You see: Add the manifest's required credentials before launching a space

Cause: The manifest needs a credential that is not set.

Fix: Add the credential on the Connections page, then bind it in the manifest. See Credentials and connections.

Space cannot start: credential unavailable

You see: This space cannot start: credential "<name>" is not available to you.

Cause: The manifest points at a credential you cannot use. The owner may have stopped sharing it, or it belongs to another person or scope. When several are missing, the message reads credentials "a" and "b" are not available to you.

Fix: Pick one of these.

Ask the credential owner to share it with you again.
Bind a credential that you own.
Ask an admin to move the credential to the team or the organization. An organization manifest needs credentials that the organization owns.

The manifest cannot launch a space

You see: Manifest can't launch a space yet

Cause: Marshal checked the manifest before launch and found a problem. The list under the heading says what to fix. One common line is version "X" is not in the available version list. A stack keeps its own copy of each package’s versions, and the version you picked is not in that copy.

Fix: Fix the items in the list. For the version line, an admin opens the stack in the Stacks page, chooses Upgrade all, and publishes a new stack version. Then point the manifest at the new version.

Launch is refused: compute class not allowed

You see: A message that starts with compute class:, followed by one of:

compute class is not in this org's allowed set
compute class is not in this stack's allowed set
compute class "<name>" is disabled

Cause: The manifest asks for a compute class that your org or the stack does not allow, or that an admin turned off. A class must be allowed by both.

Fix: In the manifest editor, pick a compute class from the list it offers. Those are the classes your org and stack allow. An admin can change the allowed classes in Settings, Compute. See Compute sizes.

A space fails to start

You see: Space failed to start

Cause: A bootstrap step failed. The page lists each step and shows the error of the step that failed.

Fix: Read the error under the failed step. Most are one of the entries on this page or on Credentials and integrations. If the error names a host that is blocked, see Network and egress.

A space shows an error

You see: Space encountered an error

Cause: The space is in the error state. The page tells you whether the workspace is safe.

Fix: If you see a Start button, the last start failed but your workspace is safe on its saved volume. Press Start to try again. If there is no Start button, read the error shown under the message, then delete the space and create a new one.

A package setup command was skipped

You see: The space page shows 1 non-fatal issue during bootstrap, with a creds line that starts bootstrap refused:.

Cause: Some packages run a setup command before the agent starts, for example to sign a CLI in with its credential. Marshal runs these commands as the agent user and keeps the state they create, which can hold the credential, out of the agent’s reach. A command that would keep that state in the home directory or in /tmp, or prepare a directory outside the places Marshal allows, is skipped. The line names the command, the rule and the path. The space starts and every other command runs.

Fix: Ask your admin to install the package’s latest version and publish a new stack version. Then select Upgrade to in your manifest and start a new space. If your team maintains the package, follow the rule the line names: keep CLI state in one directory under /run/credentials.

A toolchain runs as a different version

You see: Space created. Some saved toolchain versions run as another version.

Cause: The manifest asks for a toolchain version that has no ready-made image. Marshal started the space with another version.

Fix: Check which version the space runs. To get the exact version, an admin turns on Allow On-Demand Toolchain Downloads in Settings, Configuration, and you create a new space.

A downloaded toolchain version cannot start

You see: The space page shows a toolchains line in the bootstrap issues, in the form <toolchain> <version> cannot start: <program> needs the shared library <library>, which is not installed in this space.

Cause: Marshal downloaded the toolchain version the manifest asks for and test-started it. That build of the toolchain needs a system library the space’s image does not include. The space starts, and the line names the toolchain, the version, the program and the library.

Fix: Choose another version of the toolchain in the manifest, for example the newest version that Marshal offers ready-made, and create a new space.

Start asks you to choose how to resume

You see: Starting a space shows This size change moves the workspace, or the CLI or an agent reports that the compute class now keeps the workspace on ephemeral storage.

Cause: An admin changed the space’s size from a persistent disk to node-local storage. Resuming on the new size would move the workspace off its persistent disk, so Marshal asks first and changes nothing until you choose.

Fix: Start the space in the web app and choose Resume on the new size to take the change, or Resume as it is to keep the space’s current size and its persistent disk. See Compute sizes.

Resize failed

You see: A Resize failed message from the Resources tab, with one of these reasons:

Reason	Cause	Fix
`live resize is not enabled for this space's compute class`	The compute class has its resize policy set to off.	An admin turns on a resize policy for the class in Settings, Compute. Or change to another class; it applies at the next restart.
`... exceeds this class's resize ceiling of ...`	You asked for more CPU or memory than the class’s ceiling.	Choose a smaller step, or move to a larger compute class.
`... is below this space's cpu request of ...` (or `memory request`)	A limit cannot go below what the space requests.	Choose a value at or above the request.
`this space's pod is Guaranteed (its limits equal its requests)...`	A Guaranteed class keeps its exact size while the space runs.	Change the compute class to one where at least one limit is above its request, then restart the space.
`cpu_limit and memory_limit both equal this space's requests...`	The change would make the space Guaranteed, which cannot happen while it runs.	Keep at least one limit above its request.

See Compute sizes.

Space missing from the Spaces list

You see: A space you created is no longer on the Spaces page.

Cause: It may be archived. Archived spaces are on the Archive page. Each shows a countdown such as 3d left, and Deleting soon on the last day. The length of the window is set by your admin.

Fix: Open the Archive page and choose Restore. From the CLI, run marshal spaces restore <id>, then marshal spaces resume <id>. A restored space comes back in the stopped state. After the countdown ends, the space and its workspace are deleted for good.