Snowflake Outage Caused by Schema Update, Affected Multiple Regions Worldwide

==========================================================================

This week, Snowflake experienced a significant global outage due to a recent software update, impacting many users' ability to query data, ingest files, and causing error messages for approximately 13 hours. The company disclosed that the incident was triggered by a backwards-incompatible database schema update introduced in the latest release.

Details of the Outage

According to Snowflake’s incident report, the problematic update caused previous release packages to reference updated fields erroneously, leading to version mismatch errors and operational failures. The outage affected 10 of Snowflake’s 23 international regions, including:

• US regions: data centers in Virginia (Azure) and Oregon (AWS)
• Global regions: Ireland, Zurich, London, Sweden, Mumbai, Singapore, and Mexico

The failure was first reported early Tuesday at 0255 UTC, when customers encountered SQL execution errors during data queries. Snowflake acknowledged the issue roughly an hour and a half later and confirmed they were working on a solution. System restoration was achieved by approximately 0500 UTC.

Customer Reactions and Follow-up

Users expressed frustration on social platforms; one user commented on Reddit, “Took them way too long to roll back this change.” Snowflake committed to releasing a comprehensive root cause analysis within five days, by Sunday, though a spokesperson stated, “We have nothing additional to share at this time.”

Previous Incidents and Industry Context

This incident marked Snowflake’s second major disruption within a week, following a December 10 infrastructure issue affecting its AWS Oregon data center, which led to degraded performance.

Snowflake is not the only player facing outages recently. Between December 11 and 13, Databricks experienced a multi-day service disruption, particularly impacting its Mosaic AI assistant and users in seven US regions on Azure. Additionally, Databricks’ US Gov West AWS region suffered a complete two-hour outage on December 3.

Unlike Snowflake, which publicly details outage causes and publishes root cause analyses, Databricks has not disclosed the reasons behind their recent outages and did not respond to comment requests.