How can adtechs recover from aggregations failures?

[palenica]

How can adtechs recover from failures (such as a misconfigured batch)?

[ruclohani]

Aggregation service jobs can fail in certain scenarios e.g. job misconfiguration, coordinator service outages, etc.
Adtech should use the output of getJob API to look at the return code, message and error summary.

If the job is misconfigured, the aggregation service will provide details on error resolution in the return message. If the job succeeds partially, with some invalid reports, the response error summary will have more details about the report error counts with corresponding description. A failed job can be rescheduled using the createJob API after resolving issues (if any) or in case of any transient issues.

In certain rare cases, a failed job may result in aggregatable report accounting budget consumption without getting summary results. We are exploring solutions to help recover accidently consumed aggregatable report accounting budget in such scenarios. In the meantime, we advise ad techs to generate and keep debug reports, since processing them aren’t limited by aggregatable report accounting budget (see https://github.com/privacysandbox/aggregation-service/blob/main/docs/debugging.md).

[keke123]

We recently launched a budget recovery solution that can be used by adtechs to recover from budget-related failures as described above. Details of the criteria and process can be found here: https://github.com/privacysandbox/aggregation-service/wiki/Aggregate-Report-Accounting-Budget-Recovery-Criteria-&-Process

We welcome adtechs to try out the feature and provide feedback on the criteria and process.