Google Cloud us-west1 outage: fiber work broke Compute Engine, GKE and more

Google Cloud’s us-west1 region, in Oregon, had a broad outage on August 20, 2026. Google’s incident report says services saw timeouts, errors, higher latency and failed provisioning from 08:00 to 10:22 US/Pacific, about 2 hours and 22 minutes. Google published a preliminary report on August 24 and a full incident report on August 27, 2026.

What broke

The report lists more than 20 affected products, including:

• Compute Engine, Persistent Disk and Google Kubernetes Engine
• Cloud Run, App Engine and Cloud Build
• Cloud SQL, AlloyDB, Cloud Bigtable and Cloud Storage
• Cloud Pub/Sub, BigQuery and Identity and Access Management
• Cloud Filestore, Cloud Key Management Service and Cloud Monitoring

Some services took longer to recover because of backlogs and control-plane state that had to be reconciled. The report names Compute Engine VM provisioning in zone us-west1-c and Cloud Filestore instance allocation as examples. The header of the status page shows a wider window, 08:40 to 12:20 PDT, than the report’s summary.

Why it happened

According to Google, the trouble started during scheduled fiber optic maintenance, which unexpectedly reduced network capacity between data centers inside the region. Automated rerouting then failed to move traffic onto other capacity. The resulting congestion caused packet loss, throttling and latency, which spread into timeouts and errors in services that depend on the network.

How Google fixed it

Engineers drained traffic away from the affected network. Once fiber capacity between the campuses was restored, they moved traffic back into us-west1 in stages. Google says it is working on stricter safety checks for maintenance, redundancy requirements for circuits and better automated failover within the region. The report says these steps are still being finalized and gives no dates.

What to do

If your servers, containers or databases live only in us-west1, this outage hit every layer at once: compute, storage and identity. Keep backups and a tested restore path in a second region, and check whether your Google Cloud SLA entitles you to service credits for the time affected.