Cx Infra

Infrastructure Resources

All commands live under cx infra resources and are read-only. Run cx infra resources <command> --help for flags.

"Infrastructure monitoring is not enabled for this team" means the team has no infrastructure data. Stop and say so. Do not retry or change filters.

A stderr line starting warning: ... could not verify that infrastructure monitoring is enabled is informational: the enablement check itself failed and the command still ran.

Resource ids

• Take them from list (resource_id) and pass them as-is, quoted.
• health-history, config-changes and config-diff take up to 100 ids per call. Pass them all in one call, not one call each. Over 100, the CLI refuses. Split into batches.
• An id belongs to one team, so the id commands refuse more than one -p. Only types, filters and list fan out across profiles. From a multi-profile list, use the row's profile to pick the one to query.
• Ids mean nothing outside this skill. To reach other skills, use the resource name or its Service value.

Finding resources

1. filters first. Attribute names and values differ per type, so never guess. Per attribute: kind (string is free text, status a fixed set), values (the accepted set) and wildcard (whether * works).
2. list with the filters:
   • Every --match-all must match, at least one --match-any must match, and the two groups are ANDed. --match-all OS=linux --match-any Health=Critical --match-any Region=eu-west-1 means OS=linux AND (Health=Critical OR Region=eu-west-1).
   • Commas give one attribute several values. In --match-any any one matches. In --match-all all must, which only makes sense for wildcards: --match-all 'Name=*alert*,*processing*'.
   • Name an attribute once, in one flag. A repeat is refused.
   • --category and --type are plain filters, not requirements, and are not attributes (no --match-all Category=...). At least one filter or scope flag is required.
   • Exact values are case-sensitive (OS=linux, not OS=Linux). Wildcards are not. Quote wildcards so the shell leaves * alone.
   • --name-filter and --scope (service, environment, team) are legacy flags. They need both --category and --type, and cannot mix with --match-*. --match-all 'Name=*web*' does what --name-filter web does.
3. Zero rows is an answer. Don't retry. A wrong attribute or value is refused with the accepted ones listed.

list -o json is an envelope: {total_count, returned_count, resources}. The other commands return plain arrays or objects.

• It returns one window, rows 0 to 100 by default (--start-row, and --end-row which is exclusive). Keep paging while start_row + returned_count < total_count.
• A window cannot pass row 10,000. For bigger results, narrow the query.
• With several profiles, each one pages against its own total_count (see counts_by_profile).

Health

• list rows carry health_policies: [{id, name, status}] with status one of healthy, critical or pending (not evaluated yet). A critical policy says why a resource is critical. No extra call needed.
• health-history returns [{resource_id, health_history: [{timestamp, status}]}] with status one of Healthy, Critical or Unmonitored. The two sets of statuses have different casing.
• Ids that don't parse, and repeats, are dropped. A shorter answer is normal and stderr gives the count. Returned ids are normalized, so they may not string-match the ones sent.
• Infra health is not Service Catalog health. Correlate it with telemetry rather than treating the two as the same.

cx infra resources list --match-all Health=Critical -o json \
  | jq '.resources[] | {name, failing: [.health_policies[] | select(.status == "critical") | .name]}'

cx infra resources health-history "1001234:host_id=i-abc" "1001234:host_id=i-def" -o json \
  | jq '.[] | {resource_id, critical: [.health_history[] | select(.status == "Critical") | .timestamp]}'