Chris Baker created MESOS-6787:
----------------------------------
Summary: WebUI does not browse sandbox correctly for re-used tasks
Key: MESOS-6787
URL: https://issues.apache.org/jira/browse/MESOS-6787
Project: Mesos
Issue Type: Bug
Components: webui
Affects Versions: 1.0.1
Reporter: Chris Baker
Priority: Minor
Using dcos-vagrant with DC/OS 1.8.6 (Marathon 1.3.2, Mesos
1.0.1/d5746045ac740d5f28f238dc55ec95c89d2b7cd9)...
Marathon, when using a persistent volume, apparently reuses the same Mesos
task. Restart results in multiple "runs" of this task. In the Mesos working
directory, each run gets its own sandbox directory, under the runs directory
for the task. For example:
.../slaves/1f60eecc-a398-454d-b006-c474d9158b03-S4/frameworks/1f60eecc-a398-454d-b006-c474d9158b03-0000/executors/some-task.721ed182-c0b3-11e6-9faf-70b3d5800001/runs
The WebUI lists all of these runs in the table of Active/Completed tasks
(though, it does not indicate the UUID for the run). However, the Sandbox
hyperlink takes a user to the sandbox for the same run, regardless of which
table entry the user clicks (there doesn't seem to be anything in the WebUI
that acknowledge distinct runs). And the breadcrumbs on the file browser do not
allow to climb up to the runs directory and down to a specific run. So, without
directly accessing the agent working dir (via console or the REST API), the
other runs are shadowed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
