How can I retrieve logs from a failed replica in SkyServe?
I'm using SkyPilot 0.5.0 with AWS integration and autoscaling. When a replica fails to start, I'm unable to find logs or events related to the failure. How can I retrieve logs from a failed replica in SkyServe?
Bence Liszkai
Asked on Apr 03, 2024
To retrieve logs from a failed replica in SkyServe, you can use the sky serve logs
command with the service name and replica ID. This command allows you to stream the logs of a specific replica, providing insights into why the replica failed to start. Additionally, running sky launch service.yaml
to launch a single replica can help troubleshoot before scaling up with sky serve up
. Adjusting the initial_delay_seconds
in the readiness_probe
configuration can also be beneficial for services requiring longer initialization time.