[
https://issues.apache.org/jira/browse/AURORA-1784?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
John Sirois updated AURORA-1784:
--------------------------------
Description: We currently run the {{mesos-slave}} with
{{\--image_provisioner_backend=copy}} inside the Vagrant vm. This can result
in task launch failures due to slow copies as documented in
https://issues.apache.org/jira/browse/MESOS-6004. We currently work around
these failures by launching the {{mesos-slave}} with
{{\--executor_registration_timeout=1mins}}, but it would be better to use a
more performant means of provisioning tasks' images. One option is to flatten
e2e test images and try {{\--image_provisioner_backend=bind}}. Another is to
investigate the storage drive used by the docker daemon running in the Vagrant
vm and switch it out for a more performant driver if one can be found. (was:
We currently run the {{mesos-slave}} with {{--image_provisioner_backend=copy}}
inside the Vagrant vm. This can result in task launch failures due to slow
copies as documented in https://issues.apache.org/jira/browse/MESOS-6004. We
currently work around these failures by launching the {{mesos-slave}} with
{{--executor_registration_timeout=1mins}}, but it would be better to use a more
performant means of provisioning tasks' images. One option is to flatten e2e
test images and try {{--image_provisioner_backend=bind}}. Another is to
investigate the storage drive used by the docker daemon running in the Vagrant
vm and switch it out for a more performant driver if one can be found.)
> e2e tests fail in http_example_unified_appc due to slow image copies
> --------------------------------------------------------------------
>
> Key: AURORA-1784
> URL: https://issues.apache.org/jira/browse/AURORA-1784
> Project: Aurora
> Issue Type: Bug
> Components: Docker, Testing
> Reporter: John Sirois
> Assignee: John Sirois
>
> We currently run the {{mesos-slave}} with
> {{\--image_provisioner_backend=copy}} inside the Vagrant vm. This can result
> in task launch failures due to slow copies as documented in
> https://issues.apache.org/jira/browse/MESOS-6004. We currently work around
> these failures by launching the {{mesos-slave}} with
> {{\--executor_registration_timeout=1mins}}, but it would be better to use a
> more performant means of provisioning tasks' images. One option is to
> flatten e2e test images and try {{\--image_provisioner_backend=bind}}.
> Another is to investigate the storage drive used by the docker daemon running
> in the Vagrant vm and switch it out for a more performant driver if one can
> be found.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
