Extending workspace timeouts fails when workspaces are running in multiple regions

[andrew-farries]

Bug description

If I have workspaces running in multiple regions (eg one in the US and one in the EU), an attempt to extend the workspace timeout in either workspace will fail.

Steps to reproduce

• Create a workspace in the US region
• Create a workspace in the EU region
• Attempt to extend the workspace timeout in the EU workspace

The attempt to extend the timeout fails with an error like:

Workspace affected

No response

Expected behavior

It should be possible to extend a workspace timeout regardless of which region my other workspaces are running in.

Example repository

No response

Anything else?

No response

[andrew-farries]

The cause is here:

gitpod/components/server/ee/src/workspace/gitpod-server-impl.ts

Lines 308 to 327 in b459f3a

	// if any other running instance has a custom timeout other than the user's default, we'll reset that timeout
	const client = await this.workspaceManagerClientProvider.get(runningInstance.region);
	const defaultTimeout = await this.userService.getDefaultWorkspaceTimeout(user);
	const instancesWithReset = runningInstances.filter(
	(i) => i.workspaceId !== workspaceId && i.status.timeout !== defaultTimeout && i.status.phase === "running",
	);
	await Promise.all(
	instancesWithReset.map((i) => {
	const req = new SetTimeoutRequest();
	req.setId(i.id);
	req.setDuration(this.userService.workspaceTimeoutToDuration(defaultTimeout));
	return client.setTimeout(ctx, req);
	}),
	);
	const req = new SetTimeoutRequest();
	req.setId(runningInstance.id);
	req.setDuration(this.userService.workspaceTimeoutToDuration(duration));
	await client.setTimeout(ctx, req);

When extending the workspace timeout, we also attempt to extend the timeout in all other running workspaces (why?), but we do so with the wrong client - we reuse the one for the region of the workspace where we requested the extension.

[geropl]

Thanks for investigating @andrew-farries ! 💪 Moved to scheduled. 📓