API Timeouts

What to do

• Check whether you use Bland AI API. Check your stack

  Paste a package.json, lockfile, Terraform or a list of domains at down.guide/stack to see every provider you depend on.

• Confirm your retries, queues and scheduled jobs caught up after the 1h 36m it lasted.

  Anything that failed between 2026-04-13 13:34 UTC and 2026-04-13 15:10 UTC may need a replay.

• Subscribe to Bland AI's feed to hear about the next one first. RSS feed

  https://down.guide/feed?provider=bland works in Slack (/feed subscribe), Discord feed bots and any RSS reader.

Updates from Bland AI 4

1. investigatingApr 13, 2026, 1:34 PM UTC

   We are observing heightened API timeouts, affecting both inbound and outbound calling. We are actively investigating the alarms.

   provider

2. monitoringApr 13, 2026, 2:22 PM UTC

   Error rate has decreased, calls and dashboard access are operational again. Continuing to monitor as we investigate root cause.

   provider

3. resolvedApr 13, 2026, 3:10 PM UTC

   API is stable, a root cause at the database level has been identified. Once confirmed, postmortem will be updated with findings.

   provider

4. postmortemApr 13, 2026, 5:38 PM UTC

   # **API Service Disruption — April 13, 2026** **Resolved** | Impact: 06:34–07:22 PDT \(48 minutes\) | Severity: Major ## **What happened** For approximately 48 minutes on the morning of April 13, our API was degraded. Most affected requests timed out on the client side after 60 seconds or returned HTTP 5xx errors; voice-call orchestration, SMS send, and webhook callbacks were the most visibly impacted. The service recovered on its own at 07:22 PDT. No data was lost. ## **Cause** A chronic database condition had been accumulating silently for six days: a background replication process prevented our primary Postgres database from reclaiming space left by updated rows, and our busiest update-heavy table gradually developed significant bloat. On the morning of April 13, the resulting slowdown was enough to saturate our API-side connection pools. API workers blocked waiting for the database, new requests queued behind them and hit the 60-second timeout, and the backlog took 40 minutes to drain. ## **What we're doing** * Cleared the stuck replication state and scheduled database maintenance on the affected tables. * Adding monitoring to catch this chronic condition days in advance. * Adding an API-level circuit breaker so a slow database cannot cascade into a stopped API again. * Rewriting the slow query patterns that reduced our operational headroom. ‌ We're sorry for the disruption. If this incident affected your workload and you'd like a more detailed discussion, please reach out to your account contact.

   provider

Corrections

None so far.

Something wrong on this page? Report an error or send us a note. Every fix is listed here.