Google Cloud Status Dashboard

This page provides status information on the services that are part of Google Cloud Platform. Check back here to view the current status of the services listed below. For additional information on these services, please visit cloud.google.com.

Google BigQuery Incident #18018

Streaming API issues with BigQuery

Incident began at 2016-07-25 17:03 and ended at 2016-07-25 18:14 (all times are US/Pacific).

Date Time Description
Jul 27, 2016 16:06

SUMMARY:

On Monday 25 July 2016, the Google BigQuery Streaming API experienced elevated error rates for a duration of 71 minutes. We apologize if your service or application was affected by this and we are taking immediate steps to improve the platform’s performance and availability.

DETAILED DESCRIPTION OF IMPACT:

On Monday 25 July 2016 between 17:03 and 18:14 PDT, the BigQuery Streaming API returned HTTP 500 or 503 errors for 35% of streaming insert requests, with a peak error rate of 49% at 17:40. Customers who retried on error were able to mitigate the impact.

Calls to the BigQuery jobs API showed an error rate of 3% during the incident but could generally be executed reliably with normal retry behaviour. Other BigQuery API calls were not affected.

ROOT CAUSE:

An internal Google service sent an unexpectedly high amount of traffic to the BigQuery Streaming API service. The internal service used a different entry point that was not subject to quota limits. Google's internal load balancers drop requests that exceed the capacity limits of a service. In this case, the capacity limit for the Streaming API service had been configured higher than its true capacity. As a result, the internal Google service was able to send too many requests to the Streaming API, causing it to fail for a percentage of responses.

The Streaming API service sends requests to BigQuery's Metadata service in order to handle incoming Streaming requests. This elevated volume of requests exceeded the capacity of the Metadata service which resulted in errors for BigQuery jobs API calls.

REMEDIATION AND PREVENTION:

The incident started at 17:03. Our monitoring detected the issue at 17:20 as error rates started to increase. Our engineers blocked traffic from the internal Google client causing the overload shortly thereafter which immediately started to mitigate the impact of the incident. Error rates dropped to normal by 18:14.

In order to prevent a recurrence of this type of incident we will enforce quotas for internal Google clients on requests to the Streaming service in order to prevent a single client sending too much traffic. We will also set the correct capacity limits for the Streaming API service based on improved load tests in order to ensure that internal clients cannot exceed the service's capacity.

We apologize again to customers impacted by this incident.

SUMMARY:

On Monday 25 July 2016, the Google BigQuery Streaming API experienced elevated error rates for a duration of 71 minutes. We apologize if your service or application was affected by this and we are taking immediate steps to improve the platform’s performance and availability.

DETAILED DESCRIPTION OF IMPACT:

On Monday 25 July 2016 between 17:03 and 18:14 PDT, the BigQuery Streaming API returned HTTP 500 or 503 errors for 35% of streaming insert requests, with a peak error rate of 49% at 17:40. Customers who retried on error were able to mitigate the impact.

Calls to the BigQuery jobs API showed an error rate of 3% during the incident but could generally be executed reliably with normal retry behaviour. Other BigQuery API calls were not affected.

ROOT CAUSE:

An internal Google service sent an unexpectedly high amount of traffic to the BigQuery Streaming API service. The internal service used a different entry point that was not subject to quota limits. Google's internal load balancers drop requests that exceed the capacity limits of a service. In this case, the capacity limit for the Streaming API service had been configured higher than its true capacity. As a result, the internal Google service was able to send too many requests to the Streaming API, causing it to fail for a percentage of responses.

The Streaming API service sends requests to BigQuery's Metadata service in order to handle incoming Streaming requests. This elevated volume of requests exceeded the capacity of the Metadata service which resulted in errors for BigQuery jobs API calls.

REMEDIATION AND PREVENTION:

The incident started at 17:03. Our monitoring detected the issue at 17:20 as error rates started to increase. Our engineers blocked traffic from the internal Google client causing the overload shortly thereafter which immediately started to mitigate the impact of the incident. Error rates dropped to normal by 18:14.

In order to prevent a recurrence of this type of incident we will enforce quotas for internal Google clients on requests to the Streaming service in order to prevent a single client sending too much traffic. We will also set the correct capacity limits for the Streaming API service based on improved load tests in order to ensure that internal clients cannot exceed the service's capacity.

We apologize again to customers impacted by this incident.

Jul 25, 2016 19:18

We experienced an issue with BigQuery streaming API returning 500/503 responses that has been resolved for all affected customers as of 18:11 US/Pacific. We will conduct an internal investigation of this issue and make appropriate improvements to our systems to prevent or minimize future recurrence. We will provide a more detailed analysis of this incident once we have completed our internal investigation.

We experienced an issue with BigQuery streaming API returning 500/503 responses that has been resolved for all affected customers as of 18:11 US/Pacific. We will conduct an internal investigation of this issue and make appropriate improvements to our systems to prevent or minimize future recurrence. We will provide a more detailed analysis of this incident once we have completed our internal investigation.