Closed gruyaume closed 6 months ago
looks like this crash is happening in collect-status
right after the tls integration is gone. My initial impression is that when tls-certificates-access-relation-changed
fires, the workload is restarted, and collect status immediately tries to connect to a workload that's restarting. This doesn't explain how it may sometimes work. Maybe it's happening so close to the pebble call that an http request gets through before k8s/pebble could schedule a restart.
Connection Refused is being caught at the client level right now, but it seems like python is actually throwing a requests.exceptions.ConnectionError
to the client code. This is the exception that should probably be caught in Vault.authenticate
.
Describe the bug
An error during the
tls-certificates-access-relation-changed
causes the charm to go to an error state. This error is sometimes observed during the integration tests with the test namedtest_given_vault_deployed_when_tls_access_relation_created_then_existing_certificate_replaced
.To Reproduce
Expected behavior
No error
Logs
Juju status
Juju debug-logs
Environment
main
(commitc9a1d95978e219cc6193b6289436a7e6797720c2
)juju --version
): 3.4kubectl version --short
): MicroK8s 1.29-strict/stable