Cases2 cases

1 to 2 of 2

1. IL-0257 canva.com went down for about 50 minutes when 1.5 million requests per second OOM-killed every API Gateway task

   Canva2024

   Symptoms

   On 12 November 2024, after a new version of Canva's editor was deployed at 8:47 AM UTC, the editor's object panel stayed in a perpetual loading state for many users, particularly in Asia. Then from 9:08 AM UTC to about 10:00 AM UTC canva.com was unavailable and all requests to it failed.

   Service outageOOMKilledHigh latencyPacket loss

   Trigger

   At 8:47 AM UTC a new version of the Canva editor page was deployed, and client devices began fetching its new static assets through Cloudflare while the network path between Cloudflare's Singapore and Ashburn locations was losing packets.

   Root cause

   Dependency failure

   Component

   CDN and edge networkAPI gatewayAmazon ECS

   Contributing factor

   Overlapping failuresMonitoring blind spotRetry stormAutomation amplificationUnpatched known bug

   Remediation

   • Added a temporary Cloudflare firewall rule at 9:29 AM UTC that blocked all traffic at the CDN so new API Gateway tasks could start
   • Redirected canva.com to the status page
   • Restored traffic gradually, starting with Australian users under strict rate limits

   Action items

   • Write a runbook for rerouting, blocking and progressively restoring traffic
   • Show users a more informative error page when canva.com is unavailable
   • Raise the baseline number of API Gateway tasks and the memory of each task
   • Add load shedding rules to the API Gateway for similar traffic patterns
   • Load test the API Gateway regularly
   • Deploy the patch for the thread-locking bug in the telemetry library
   • Add multithreaded tests for the contended code path to the telemetry library microbenchmark suite
   • Use page load completion events as a second canary indicator for page asset releases
   • Experiment with adding another rollout stage to the canary to give more time to detect issues
   • Add timeouts so user requests do not wait excessively long for assets
   • Work with Cloudflare to understand the system interactions behind the incident

   View case
2. IL-0461 Every rollout of Datadog's metrics query service on Kubernetes raised errors that the logs blamed on DNS

   Datadog2022

   Symptoms

   From September 2021, each rollout of an update to Datadog's metrics query service, which serves real-time and historical metric data to the web frontend and to monitor evaluation, came with a rise in errors. Clients retried failed queries, so few errors reached users, but the retries slowed dashboards and monitors. The service's logs showed DNS errors stopping it from connecting to the data stores it depends on.

   High error rateDNS resolution failuresHigh latencyOOMKilledPacket lossStale endpoint routing

   Trigger

   Rollouts of the metrics query service, which delete the old pods. The client behaviour behind the failures came from an August 2021 rollout of the metrics query clients that applied round_robin as the new default gRPC load balancing policy.

   Root cause

   Configuration change

   Component

   conntrack tableCiliumLinux kernelCoreDNSClient-side DNS resolver and cacheexternal-dnsHTTP and RPC client library

   Contributing factor

   Change with a delayed effectRetry stormUnknown limitMonitoring blind spot

   Detection

   Errors rose during rollouts of the metrics query service, and the service's logs showed DNS errors when connecting to its data stores.

   Remediation

   • Raised the memory of the node-local-dns pods from 64 MB to 256 MB, which stopped the OOM kills but not the DNS errors
   • Tested larger instance types and confirmed that scaling up the service's infrastructure resolved the errors
   • Removed the custom gRPC reconnect parameters from the clients to restore the defaults

   Action items

   • Contribute a change to Cilium that adds an unreachable route for a deleted pod's IP so clients get an ICMP error

   Lessons

   The errors looked like DNS but were not. ENA metrics and VPC Flow Logs gave the visibility into VPC connection tracking that host metrics did not, and Cilium, Kubernetes and AWS networking interact in ways that can make reverse path filtering drop packets it should allow. DNS propagation time depends on many settings. The causes were changes Datadog had made itself, to work around a dependency bug, to resolve earlier incidents and to override default reconnect parameters, which looked safe at the time and showed their effects much later, when the link was hard to see.

   View case

By root cause2