kubectl access first. Ankra reads container logs through the Ankra Agent, and searches older lines in a log store: Ankra hosted logs, or a store of your own.
There are three places to read logs:
Reading logs needs the
kubernetes.logs permission on the cluster. The owner, admin, operator and member roles hold it; viewer does not, because logs carry whatever a workload prints. See Roles & Access.
Cluster logs
Open Infrastructure → Clusters → <your cluster> → Technical details → Kubernetes → Cluster logs (/next/clusters/<cluster-id>/kubernetes/logs). The page has two modes, switched with History and Live at the top right. A badge under the title says where the lines come from: the log store’s name, or Live from the containers.
History
History searches the lines a log store has kept. It opens on the last hour.- Type in Search text, a request ID, an error… and press
Enter, or leave it empty to see every line. - Pick a time window with the time range picker and click Search. A window covers at least a minute and at most 30 days, and ends no later than now.
- Narrow with Filters: Namespace, Pod, Container and Severity (each level, plus No level). Each filter you add shows as a chip in the search box; click its × to remove it.
When the cluster has no log store, History shows Search this cluster’s log history with three choices: Ankra hosted logs (with Turn on hosted logs if your role can change the cluster), Your own log store (Connect a log store opens Settings → Logs and data, which links to Integrations, where the store is connected), and Live logs, right now (Read live logs).
Live
Live reads what the containers of one namespace are writing, straight through the agent. No log store is needed, but you only get what the containers still hold.- Pick a namespace in Choose a namespace to read live, or in the Namespace field.
- Choose how far back each container starts in Lines: Last 100 per container, Last 300 per container (the default) or Last 1,000 per container.
- Type in Filter lines as you type… to narrow the loaded lines, and use Filters for Pod, Container and Severity.
Live · <pods shown> of <pods> pods · <lines> lines. Pause stops new lines arriving, and Resume starts them again. Previous run shows what each container wrote before its last restart, which is where a crash-looping container’s error is; only a container that restarted has one.
Read a line
Click a line to open Log line: its time in UTC, namespace, pod and container (each with a button for only lines from it), the message, any structured fields, the raw line and the lines around it. Copy line copies it and Copy link copies an address that opens the same line selected.Lines and patterns
Above the results, Lines shows each line and Patterns groups lines with the same shape, with the numbers, IDs and addresses in them set aside, and counts them. A flood of health checks becomes one row, and the one new error stands out. Click a pattern to see its lines. Order switches between Newest first and Oldest first, and Wrap long lines wraps them. To save what you see, click Export the shown lines as NDJSON: one JSON object per line, forjq or another tool.
Ask Ankra about the logs
When lines are shown, Ask Ankra at the top asks Ankra AI what is going wrong in them. The question carries the cluster, the window or live view, the namespace and pod filters and the search text, and the count of errors and warnings among the shown lines, with the three most frequent error patterns.Keyboard shortcuts
Press? on the page for the list.
A pod’s Logs tab
Open the pod (Kubernetes → Workloads → Pods, then the pod, or View logs on its row) and click Logs.- Pick the container in Container.
- Choose Current instance, or Previous instance for the run before the last restart.
- Choose Last 100 lines, Last 500 lines (the default) or Last 1,000 lines.
- Click Follow to read the tail again every five seconds; Pause following stops it.
From the CLI
ankra cluster logs streams a pod’s logs, like kubectl logs -f. -n is required.
[pod] or [pod/container]. See ankra cluster logs for every flag.
AI agents connected over MCP use get_pod_logs for one pod and query_logs to search the cluster’s log store. See MCP Tools.