Troubleshooting
Quick checks before contacting support.
Always include the alert ID
Every alert has a unique uniqueEventId (also called eventId). Include it in every support request: it lets us trace the alert through each step.
No alert received
- Was the threshold crossed? Compare the move with the thresholds. Volatile assets need bigger moves.
- Was it a single-day move, during regular hours? Multi-day drifts and extended-hours moves do not trigger alerts.
- Did the same alert type already fire today for this asset?
- Is the asset in your universe, and the alert type in your feed?
- Is your consumer running? Check your consumer group lag on the topic.
No alerts at all
- Check that your consumer is connected and that your credentials have not changed.
- Check you are reading the right topic: production topics have no
stage-prefix. - On weekends and market holidays, only crypto and forex alerts are generated.
Messages fail to parse
- One-byte messages (for example
0x08) are Kafka control records. Skip records with a 4-byte key and a 6-byte value. - Read the schema from the Schema Registry by subject, not by a fixed schema ID.
- With a deserializer, the message value is already parsed. Don't decode it again.
See Kafka best practices.
REST API errors
| Status | Likely cause |
|---|---|
401 | Missing token |
403 | Invalid or expired token, or your plan does not include this data |
400 | Invalid parameter, for example an unsupported language code |
Tokens are valid 24 hours. See Errors.
Reasoning looks wrong
Send support the uniqueEventId, the language, and what is wrong (wrong company, wrong direction, old fact, translation). Every generation is traced, so we can see exactly which source and check led to the reason.
Personalized alert not received by a user
- Was the user's portfolio sent before the alert fired?
- Does the user have
productsincluding1? - Was each asset resolved? Unknown identifiers are skipped silently.
- Does the user's
PreferredLanguagesuse thexx-XXformat (for exampleen-US)?