Support Bundle Generation¶

Audience: operators triaging incidents or escalating to the AutonomyOps support team.

What is a support bundle?¶

A support bundle is a tar.gz archive containing sanitized diagnostic information collected from the local node. It includes system information, build metadata, a redacted configuration snapshot, HA status, recent audit events, and log extracts.

Secret redaction is automatic and mandatory:

fleet_salt values are replaced with <REDACTED>.
PostgreSQL URLs have their passwords replaced with REDACTED in a URL-safe manner.
No YAML parsing is required — redaction is applied via line scan with an explicit allow-list, so unrecognized secret patterns are not silently included.

Partial failure is tolerated: if one collector fails (e.g. HA status is unreachable), the bundle continues and records the error in manifest.json.

1. Generate a support bundle¶

Minimal (local node info only)¶

autonomy support-bundle generate \
  --output /tmp/support-bundles/autonomy-bundle-$(date -u +%Y%m%d-%H%M%S).tar.gz

Full bundle (with HA status and audit events)¶

autonomy support-bundle generate \
  --output /tmp/support-bundles/autonomy-bundle-$(date -u +%Y%m%d-%H%M%S).tar.gz \
  --config-file /etc/autonomy/config.yaml \
  --orchestrator-url "$AUTONOMY_ORCHESTRATOR_URL" \
  --audit-dir "$AUTONOMY_AUDIT_DIR" \
  --log-file /var/log/autonomy/edged.log

Expected output:

generating support bundle: /tmp/support-bundles/autonomy-bundle-20260318-103000.tar.gz
  collecting system_info...
  collecting build_info...
  collecting config...
  collecting ha_status...
  collecting audit_recent...
  collecting logs...
bundle written: /tmp/support-bundles/autonomy-bundle-20260318-103000.tar.gz

Example with a collector failure (HA unreachable):

generating support bundle: /tmp/support-bundles/autonomy-bundle-20260318-103000.tar.gz
  collecting system_info...
  collecting build_info...
  collecting config...
  collecting ha_status...
  warning: ha_status failed: GET http://localhost:8888/v1/ha/status: request failed: ...
  collecting audit_recent...
  collecting logs...
bundle written: /tmp/support-bundles/autonomy-bundle-20260318-103000.tar.gz

The collector warning does not abort the bundle. The error is recorded in the archive’s top-level manifest.json.

2. Bundle contents¶

The generated tar.gz contains:

File	Collector	Contents
`<bundle-dir>/manifest.json`	—	Bundle metadata, collector results, errors
`<bundle-dir>/system_info.json`	`system_info`	OS, hostname, architecture, Go version
`<bundle-dir>/build_info.json`	`build_info`	ADK version and Go build metadata
`<bundle-dir>/config_redacted.yaml`	`config`	Config with secrets replaced by `<REDACTED>` / `REDACTED`
`<bundle-dir>/ha_status.json`	`ha_status`	Output of `/v1/ha/status`
`<bundle-dir>/audit_recent.json`	`audit_recent`	Last 50 audit events from the audit store
`<bundle-dir>/logs/autonomy.log`	`logs`	Last 500 lines of the specified log file by default

4. Common failure scenarios¶

HA status collector fails¶

If --orchestrator-url is unreachable:

[error] ha_status: connection refused

This is non-fatal. The bundle is still useful for system_info, build_info, and logs. Verify the control-plane URL:

curl -sf "${AUTONOMY_ORCHESTRATOR_URL}/v1/health/read-ready" | jq .

Config collector skipped¶

If --config-file is not provided, the config_redacted collector is skipped and the file does not appear in the bundle. Always provide --config-file for a full bundle.

Audit collector produces empty results¶

If --audit-dir does not exist or contains no events:

The bundle still records audit_recent as ok, but the exported JSON array may be empty.

Verify the audit directory:

ls -la "$AUTONOMY_AUDIT_DIR"

Log file too large¶

The logs collector captures only the last 500 lines by default. For long-running nodes with large log files, include a separate compressed log extract:

tail -5000 /var/log/autonomy/edged.log | gzip > /tmp/edged-log-extract.gz

Known gaps¶

No automatic bundle rotation: Bundles accumulate wherever --output points. Implement a retention policy (e.g. keep last 5 bundles) manually or via cron.
No relay deadletter snapshot: The bundle does not include edgectl relay deadletter list output. Add this manually when diagnosing relay issues:
```
edgectl relay deadletter list --socket /run/edged/rpc.sock \
  >> /tmp/deadletter-snapshot.txt
```
No replication status: The bundle does not include ha replication or PostgreSQL pg_stat_replication output. Collect manually when diagnosing HA issues.
Audit collector reads from file store only: If the audit store is a future database-backed emitter, the file-based collector will need to be updated to query the database.