The key clue is this part:

text
POST /prefect/work_pools/default_pool/workers/heartbeat ... 500 ... 10000 ms
Error #01: context deadline exceeded

What it means: the worker started successfully enough to find the work pool (

GET ... 200

), but when it tried to send its heartbeat, the server did not finish processing that request within its timeout window and returned

500

. In Prefect 3.x, that heartbeat endpoint updates the worker record in the backend database. So this is usually not a worker startup problem — it is usually a server/database timeout problem. Most likely reasons 1. Database is slow or blocked - heartbeat writes to the Prefect DB - if the DB is under load, locked, or slow, the request can time out 2. Server DB timeout is too low - the exact

10000 ms

strongly suggests a 10-second backend timeout is being hit 3. Too few DB connections / connection pool exhaustion - common if several workers or API requests hit the DB at once 4. Proxy / gateway timeout behavior - you are behind OpenShift / KrakenD / ingress - even if the proxy is not the root cause, it can expose or worsen backend latency So the likely flow is:

text
worker -> proxy/route -> Prefect API -> Prefect database

The GET succeeds quickly, but the POST heartbeat requires a DB write and that write is exceeding the timeout. Why I think this is the reason -

GET "/prefect/work_pools/default_pool"

returns

200

-

POST "/prefect/work_pools/default_pool/workers/heartbeat"

fails after exactly

10000 ms

-

context deadline exceeded

is a classic timeout symptom, typically from backend request/DB execution timing out What to check first 1. Check Prefect server database health - Is Postgres reachable and healthy? - Any slow queries / locks? - High CPU / memory on DB pod? 2. Increase the Prefect server DB timeout On self-hosted Prefect server, try increasing:

bash
PREFECT_SERVER_DATABASE_TIMEOUT=30

Then restart the Prefect server. 3. Increase DB pool settings if needed If you have multiple workers:

bash
PREFECT_SERVER_DATABASE_SQLALCHEMY_POOL_SIZE=20
PREFECT_SERVER_DATABASE_SQLALCHEMY_MAX_OVERFLOW=10

4. Check OpenShift / ingress / KrakenD timeouts Make sure upstream/proxy timeout is greater than the Prefect processing time. 5. Look at Prefect API server logs and DB logs at the same timestamp Specifically around:

text
2026-06-03 10:11:23

Important note This is probably not caused by the work pool itself, because the pool lookup succeeded. The failure happens when the worker heartbeat tries to write/update worker state. Most probable root cause in one sentence: Your worker can reach the Prefect API, but the API cannot complete the heartbeat database operation within 10 seconds, so it returns

context deadline exceeded

. If you want, I can help you with the exact OpenShift env vars / Helm values to increase the timeout for your Prefect server deployment.