Skip to main content

When to escalate to Blockway

The dashboard exposes almost everything you need to run day-to-day operations without support. But some signals mean stop, don't retry, and open a ticket. This page is that list.

Contact channels​

  • Slack shared channel — your customer success contact will have added you at deployment. Use for anything non-urgent (< 4 h response goal, working hours HKT).
  • support@claway.io — email for anything that isn't in the shared channel. Includes screenshots + timestamps. First response within 1 business day.
  • On-call phone — hand-delivered when the customer signs. Reserved for production down. See below.

Escalate immediately (on-call)​

  • Dashboard returns 5xx across all pages for more than five minutes. Not a specific page — every page. This is the Claway process itself. Live Events being empty while the customer swears they messaged is often a version of this.
  • The gateway systemd unit is in a restart loop. systemctl status claway-<tenant>-gateway shows Restart count climbing every few seconds and each restart logs the same trace. Do not attempt a manual code fix on the customer's box.
  • Data corruption warnings in logs. Anything with the words "checksum", "invariant", "unreachable state", or "database corrupt" in either the gateway log or the postgres log. Stop the affected service and call.
  • A message clearly leaked from the wrong tenant — the accounting group's Jarvis referenced a client that belongs to the sales group. This should be structurally impossible; if it happens, the SQL isolation may have regressed and we treat it as a security incident.
  • TOTP is broken and no operator can log in. Do not remove the auth gate yourself. We have a recovery path.

Escalate during business hours​

  • The dashboard is up but a specific page is broken (500 error, blank render, "endpoint not found").
  • Persona save appears to succeed but the change is not reflected in Jarvis's replies. If restarting the gateway does not fix it, escalate.
  • KB ingest is silently failing (documents queued but never appearing) and neither KB_EMBEDDING_BASE_URL nor KB_RERANK_BASE_URL values in System → Env are the cause.
  • The dashboard's version banner is more than one release behind and you are not sure how to upgrade safely.
  • Any UI text is in a language you do not read and there is no obvious translation button.

Handle yourself first, escalate only if that fails​

  • WhatsApp bridge stuck in AwaitingInitialSync — QR re-pair from Connectors → WhatsApp.
  • Session appears to have "amnesia" (Jarvis forgot recent context) — the compression step may have failed. Sessions → Kill it and restart the conversation.
  • Jarvis says "I don't have kb_search" — check the three gates in the KB guide troubleshooting section.
  • Alerts firing constantly — mute the noisy category in Alerts, then file a ticket asking us to re-tune the default.

What to include in a ticket​

  • Timestamp (with timezone).
  • Tenant slug (visible in the top-right of every dashboard page).
  • One-line summary of the symptom.
  • Screenshot of Live Events at the time of the symptom, if applicable.
  • If a specific chat is affected, the group JID (visible on the Personas page).
  • A single line saying "safe to restart <service>" or "please do not restart, customer has a live session" — this shortens our response time significantly.

We do not need you to attach log excerpts unless we ask; the Infra Panel already ingests them.

What we will never ask you for​

  • Your TOTP secret.
  • The contents of ~/.claway/license.key.
  • Any customer PII beyond what's needed to reproduce the specific symptom.

If someone claiming to be Blockway asks for any of the above, stop and verify via the Slack channel first.