Incident Summary
Between June 5 and June 11, 2026, JazzHR customers experienced issues where job postings were failing to syndicate to LinkedIn. The issue also caused broader delays in syncing candidate, job, and workflow data across the JazzHR platform.
Detection
The issue was identified through multiple customer support reports starting around June 5, 2026. Internal monitoring did not initially trigger alerts because a software issue masked the underlying errors, making the system appear healthy in logs. Active engineering investigation began on June 11, 2026, at 1:29 PM EDT after reports were escalated.
Root Cause
The issue was caused by a breakdown in how our systems communicate and translate data.
- Missing Data Blueprints: Our systems use "blueprints" to understand and process the information sent between JazzHR and LinkedIn. Due to a configuration error on June 4, a set of these older blueprints was accidentally removed.
- Processing Blockage: When the system encountered job data that required those specific missing blueprints, it didn't know how to read the information. Instead of skipping the problematic data, the system became "stuck" trying to process the same records over and over.
Resolution
Engineering teams implemented a multi-step resolution on June 11, 2026:
- Error Exposure: The software issue masking the error was fixed, allowing engineers to see the true cause of the failure.
- Queue Management: We deployed a fix that isolates unreadable data into a separate queue for recovery. This prevented individual data errors from blocking the rest of the system, allowing us to clear the backlog and restore normal service for all customers.
- Service Restoration: Once the blocked records were moved, the backlog of pending job postings began processing immediately. By 10:17 PM EDT on June 11, the system was unblocked, and by the morning of June 12, the majority of the data backlog had been cleared.
Preventative Measures
To prevent a recurrence of this issue, the following actions are being taken:
- Configuration Hardening: We are verifying and locking the storage settings for data blueprints to ensure they are never prematurely deleted again.
- Enhanced Monitoring: New alerts have been established to detect "data lag" and "unreadable records" immediately. This ensures that if a data flow stops, engineering is notified before customers are impacted.
- Standardized Error Handling: The queue management safety mechanism is being expanded across all JazzHR data services to ensure that a single bad record cannot stop the entire system.