Cases2 cases

1 to 2 of 2

1. IL-0250 slack.com stopped resolving on some DNSSEC-validating resolvers for about 24 hours after DNSSEC signing was turned on

   Slack2021

   Symptoms

   Soon after Slack published the DS record for slack.com at its DNS registrar on 30 September 2021, the Customer Experience team started getting user reports of "ERR_NAME_NOT_RESOLVED" errors in Slack clients. Fewer than 1% of online users were affected, and most reports came from users of one private corporate recursive resolver and of Google Public DNS (8.8.8.8).

   DNS resolution failures

   Trigger

   On 30 September 2021 the Traffic Engineering team enabled DNSSEC signing on the slack.com zone in Amazon Route 53 and on its sub-delegated zones in NS1, and after a three-hour soak published the slack.com DS record at the DNS registrar.

   Root cause

   Software defect

   Component

   Authoritative DNS servers and recordsRecursive DNS resolver serviceTerraform

   Contributing factor

   Unrepresentative stagingUnverified assumptionHarmful manual actionIneffective rollbackLimited debug access

   Detection

   Customer Experience received user reports of "ERR_NAME_NOT_RESOLVED" errors in Slack clients, and customers sent NetLog captures through support cases. After signing was rolled back, Slack's external DNS and DNSSEC resolution probes alerted the Traffic team.

   Remediation

   • Removed the slack.com DS record from the DNS registrar
   • Rolled back DNSSEC signing on the slack.com authoritative and delegated zones, which made resolution worse
   • Asked major ISPs and public DNS resolver operators to flush their cached slack.com records
   • Flushed the slack.com records on Google's 8.8.8.8 and Cloudflare's 1.1.1.1 through their public flush sites

   Lessons

   The DNSSEC rollout had been tested only on domains without a wildcard record, so the Route 53 wildcard NSEC bug first appeared on slack.com; Slack names not testing a wildcard domain first as an oversight. Removing the DS record at the registrar does not stop validation at once, because resolvers cache the DS record from the .com zone for up to 24 hours, and removing the DNSKEY while that record is still cached makes every validating resolver fail. Managing DNS through Terraform hid the console warning about disabling DNSSEC. Re-enabling signing would neither have helped nor hurt, because Route 53 generates a new ZSK each time signing is enabled.

   View case
2. IL-0251 Most of Slack's HAProxy load balancers routed to a stale, shrinking set of webapp backends and served HTTP 503 errors

   Slack2020

   Symptoms

   At 4:45pm Pacific on 12 May 2020, Slack was alerted that it was serving more HTTP 503 errors than normal. A subset of the webapp fleet was under heavy load while the rest of the webapp instances were not.

   High error rateLoad imbalanceStale endpoint routing

   Trigger

   At the end of the US business day the webapp tier began its usual scale-down, and autoscaling terminated the oldest instances first, which were the only backends left in most HAProxy instances' state.

   Root cause

   Software defect

   Component

   HAProxyService discovery

   Contributing factor

   Automation amplificationNo alert on limit

   Detection

   An alert for more HTTP 503 errors than normal. The alerting meant to catch stale HAProxy state was not working.

   Remediation

   • Did a rolling restart of the HAProxy fleet

   Action items

   • Move all ingress load balancing from HAProxy to Envoy Proxy, with endpoint discovery through an xDS control plane

   Lessons

   The alerting for this exact failure had stopped working, and nobody noticed because the HAProxy stack had run for a long time without needing changes, so few engineers worked with its monitoring and alerting. A slot limit that had been ample a few months earlier was reached after pandemic-driven growth and one large autoscaling event.

   View case

By root cause1