When several threads are running in parallel using pselect function or Parallel macro and when of those threads raise exception
pselect exits (with exception) without waiting for other threads to complete. As a result uncaught exception causes deployment to finish while there are still running threads. Those threads can still report to UI even if no deployment is running from user perspective. Because after deployment local package cache is wiped if such thread accesses its resource file an exception (file not found) is raised and that exception is reported to UI. And if such thread will try to send execution plan that was built earlier it will restart agent listener and it will steal messages from RabbitMQ queues causing subsequent deployments hang
Steps:
1. Add any Docker Application to environment with k8s cluster
2. Deploy environment
Actual result:
Sometimes deployment fails with errors in deployment logs in dashboard:
2015-04-21 13:07:55 - Configuring etcd master node
2015-04-21 13:07:55 - Error: [Errno 2] No such file or directory: '/tmp/murano-packages-cache/8ac61cc5-5f2b-43ff-9b84-6bccd8a1a95b/io.murano.apps.docker.kubernetes.KubernetesCluster'
2015-04-21 13:07:55 - Creating Kubernetes Master
2015-04-21 13:07:55 - Creating Kubernetes Minion
2015-04-21 13:07:55 - Creating Gateway node for Kubernetes services
2015-04-21 13:07:55 - Error: [Errno 2] No such file or directory: '/tmp/murano-packages-cache/8ac61cc5-5f2b-43ff-9b84-6bccd8a1a95b/io.murano.apps.docker.kubernetes.KubernetesCluster'
Full deployment hystory: http://paste.openstack.org/show/204972/
This can be caused by http://paste.openstack.org/show/208822/ , for example.
https:/