Sign-in and access
Sources
An observer doesn’t report in
The connect dialog waits for the observer’s first report. It usually arrives within a minute. If the dialog says Taking a little longer, check that the command finished without errors. On Kubernetes, check the pod and read its logs:sudo systemctl status rk-observer on Linux. On AWS or Google Cloud installs in Docker, use sudo docker logs rk-observer.
Then check three things. The machine needs outbound access to api.rubixkube.ai:443 and nats.rubixkube.ai:4222. On Kubernetes, the node needs to be able to pull ghcr.io/rubixkube-io/observer-agent. And the key needs to be the one from your organization’s connect dialog.
A source says Disconnected
A source reads Disconnected after 5 minutes with no heartbeat. Use the checks above. If the machine was stopped or the observer was uninstalled, start it again or reinstall it.A source says Needs attention
Open the source. The page lists what’s wrong under The observer needs attention. For an observer on a big cluster, it also tells you how many pods to run.A tool isn’t sending anything
The dialog says Waiting for the first message until the tool sends something. Check these in order:- The webhook URL is in the place the dialog named, in the tool itself.
- The Datadog or Observe webhook has the payload from the dialog. Those two send whatever template they’re given.
- The automation is on. If the dialog said The automation wasn’t installed, choose Try again.
- Grafana’s contact point has Max alerts per message set to 1.
Incidents and RCAs
The RCA hasn’t started
This is usually on purpose. The incident says what it’s waiting for:
A normal wait can’t be skipped from the console. Run the RCA again shows once an RCA has finished or failed, or when none is coming. To change the waits for everyone, use Settings > Incidents.
No RCA is coming
Some incidents don’t get an RCA, and the page says why:- It recovered before the RCA started, so no RCA runs. If it breaks again, the RCA starts.
- Earlier RCAs say this is expected behaviour, so no RCA runs. This is a Known alert. Choose Ignore, or Run RCA again if it’s wrong.
- This keeps coming back. Rubix has investigated it as often as your settings allow this week. Snooze it, ignore it or run the RCA again. You can change the limit in Settings > Incidents.
- The RCA was due and has not started, or The last RCA attempt stopped. Run the RCA again, or ask Rubix what it can see.
The RCA failed
The incident says why, and when the next automatic try is. An RCA gets up to three attempts.The RCA says Not determined
Rubix didn’t have enough to state a cause, and it says so. This often means it couldn’t read something it needed. Check Customize > Integrations for what’s connected, add a comment with what you know, and use Run RCA again.Rubix
- Rubix won’t change anything. The chat is in Observe mode. Change the mode in the chat, up to the highest mode in Settings > Modes.
- You can’t pick a higher mode. An admin sets the highest mode anyone can use. See Modes.
- Rubix stopped and nothing’s happening. In Assist, it’s waiting for you. Look for the approval card in the chat.
- Rubix can’t see a tool. Open the integration under Customize > Integrations. Its Reads section shows what the connection can read. Needs attention means the connection has an error.
Slack
- Slack comes with Business. Slack needs a Business or Enterprise plan.
- Slack needs new permissions. An admin chooses Reconnect in Customize > Channels.
- Messages go to the wrong channel. Set a channel for each category in Customize > Channels. A category with no channel goes with Incidents.
MCP
Your AI tool signs in with your own RubixKube account, so sign in when it asks. The server ishttps://mcp.rubixkube.ai/mcp. See MCP.
Plans and limits
A card that says something like Automations come with Business means your plan doesn’t include that feature. Settings > Usage and plan shows your limits and marks each one Near limit, At limit or Over limit.Before you write to us
Choose Help and support in your user menu, or write to connect@rubixkube.ai. These help us answer faster:- Your Organization ID, from Settings > General
- The incident code, such as INC-12, or the source name
- The time it happened, with your timezone
- What you saw, what you expected, and a screenshot
- For an observer, the output of the
kubectlcommands above. Remove any secrets first.
FAQ
Common questions.
System status
Check whether it’s on our side.
Connect a source
Go through the install again.
Incidents
How incidents work.