This repository has been archived by the owner on Jan 30, 2020. It is now read-only.
fleetctl commands fail sporadically with "unable to find Job(...)" #582
Labels
Comments
marineam
added a commit
to marineam/coreos-overlay
that referenced
this issue
Jun 27, 2014
Rolling back to 0.5.0 due to a regression in the etcd client. Issue: coreos/fleet#582 This (mostly) reverts commit 6a01e3c.
bcwaldon
added a commit
to bcwaldon/fleet
that referenced
this issue
Jul 1, 2014
Partially fix coreos#582
crawford
pushed a commit
to crawford/coreos-overlay
that referenced
this issue
Jul 8, 2014
Rolling back to 0.5.0 due to a regression in the etcd client. Issue: coreos/fleet#582 This (mostly) reverts commit 6a01e3c.
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
The old go-etcd client was configured to make all requests with "strong" consistency. This means that all "GET" requests would be redirected the etcd leader. The new etcd client does not have the same concept of "strong" consistency. This means that all requests first hit the configured endpoint, possibly being redirected. GET requests are no longer blindly redirected to the leader.
During the
fleetctl startcommand, the Job is created then the Job's target state is set. Before the target state is set, the code attempts to GET the job. If fleetctl is not configured to talk to the leader, this GET operation may fail, as the logs are not guaranteed to have been sync'd to the local etcd endpoint.The text was updated successfully, but these errors were encountered: