Migrating Contra Costa County's Emergency Alerting Platform from CAPServer to AtHoc
Led subscriber-data migration from a legacy CAPServer alerting platform to AtHoc for Contra Costa County, keeping the live system available throughout — preceded by an onsite recovery of the outgoing platform on Christmas Eve.
Background
When AtHoc acquired Alerting Solutions, I inherited the CAPServer project along with its backlog, operational support, CAPCON support, on-call duties, and primary responsibility for Contra Costa County — one of the platform's public safety customers.
A Failing System on Christmas Eve
Shortly before the county's migration to AtHoc, CAPServer began failing. I went onsite to the Contra Costa County Sheriff's Office data center, failed the service over to the backup VM to restore service, then expanded and replaced storage on the primary system and brought it back into service. Given how close the AtHoc migration already was, the call was to stabilize the existing platform and recommend inventorying host capacity if legacy operation continued, rather than investing in unnecessary modernization of a system already being replaced.
The Migration
My largest responsibility in the AtHoc migration was the subscriber-data ETL. The source subscription system had to stay live and in use throughout — there was no maintenance window where the county's alerting subscribers could simply go dark. The migration ran as an initial bulk load, followed by delta migrations to pick up ongoing changes, with a final delta at cutover to close the gap.
The data mapping itself was straightforward: AtHoc's subscriber labels matched CAPServer's closely enough that this wasn't a complex transformation project, just a careful, live one. Validation combined record-count reconciliation with representative test alerts, comparing activation numbers between the old and new systems to confirm the migrated data behaved the same way operationally, not just that it existed.
AtHoc's engineering team built the CAPCON/AtHoc integration itself; my role there was deployment and ongoing support, not developing that integration.
Lessons
Migrating a live public-safety subscriber system reinforced that the hard part of this kind of ETL usually isn't the mapping — it's proving the migrated data works the same way in production as the data it replaced, which is why the activation-number comparisons mattered as much as the record counts. And the Christmas Eve recovery was a reminder that stabilizing a system you already know is being retired can be the right call, even when a deeper fix might be more technically satisfying.