- Joined
- Jul 14, 2016
- Messages
- 7
- Reaction score
- 1
We've hit two separate but persistent reliability issues integrating against our 3CX PBX (northpoint.3cx.us) via a client-credentials OAuth app:
1. XAPI role-claim instability under parallel requests
Firing multiple Report-API requests concurrently (e.g. one ReportExtensionStatisticsByGroup call per extension group via Promise.all) reliably produces spurious 401 Unauthorized on some — but not all — of the parallel calls, even though the exact same OAuth token and role succeed on the other calls in the same batch. Switching to sequential requests with a short retry/backoff resolves it completely. This suggests the token's effective role/permission check isn't fully consistent under concurrent load.
2. Data Connector scheduler stalls without any error surfaced
Our Data Connector (Call History/CDR, Recordings, System Metrics, Call Audio Metrics) is configured for "Every 15 Minutes," but the feed has repeatedly gone stale for periods well beyond that (once ~6 hours) with no error shown in the admin UI. Manually clicking "Transfer Data Now" immediately clears it and backfills correctly, confirming connectivity/credentials are fine — the scheduled trigger itself just isn't firing reliably. We reproduced this identically after switching the connector's target database from MySQL to PostgreSQL, so it isn't tied to the target database type.
Both are intermittent and hard to reproduce on demand, but consistent enough over the past week that we're fairly confident they're platform-side rather than something in our integration. Happy to share logs/timestamps if useful for tracking this down.
1. XAPI role-claim instability under parallel requests
Firing multiple Report-API requests concurrently (e.g. one ReportExtensionStatisticsByGroup call per extension group via Promise.all) reliably produces spurious 401 Unauthorized on some — but not all — of the parallel calls, even though the exact same OAuth token and role succeed on the other calls in the same batch. Switching to sequential requests with a short retry/backoff resolves it completely. This suggests the token's effective role/permission check isn't fully consistent under concurrent load.
2. Data Connector scheduler stalls without any error surfaced
Our Data Connector (Call History/CDR, Recordings, System Metrics, Call Audio Metrics) is configured for "Every 15 Minutes," but the feed has repeatedly gone stale for periods well beyond that (once ~6 hours) with no error shown in the admin UI. Manually clicking "Transfer Data Now" immediately clears it and backfills correctly, confirming connectivity/credentials are fine — the scheduled trigger itself just isn't firing reliably. We reproduced this identically after switching the connector's target database from MySQL to PostgreSQL, so it isn't tied to the target database type.
Both are intermittent and hard to reproduce on demand, but consistent enough over the past week that we're fairly confident they're platform-side rather than something in our integration. Happy to share logs/timestamps if useful for tracking this down.
