Second, and sometimes subsequent, network tests can fail
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Checkbox Provider - Base |
Fix Released
|
High
|
Rod Smith |
Bug Description
The change to NetPlan is causing failures of the second, and sometimes subsequent, network tests on some systems. That is, if a SUT has four network devices (let's say eno1, eno2, eno3, and eno4), the first device (eno1) may pass, but then eno2, eno3, and eno4 are likely to fail. If they're re-tested when Checkbox presents the option to do so at the end of a test run, eno2 will likely pass, but eno3 and eno4 are likely to fail; and so on. Failure messages look like this:
ERROR:root:No IP address for eno2
ERROR:root:No netmask for eno2
ERROR:root:Device eno2: Invalid IP Address
ERROR:root: Expected 4 octets in 'None'
ERROR:root:Aborting test now
The problem appears to be that the network script brings up unused devices in interface_test() via a call to "ip link... up", but doesn't wait to confirm that it's been brought up. With NetPlan, bringing the interface up can take a few seconds, so a delay -- or better, a loop to pause until the interface can be confirmed to be up with a timeout -- is necessary.
Related branches
- Maciej Kisielewski (community): Approve
-
Diff: 89 lines (+33/-4)1 file modifiedbin/network (+33/-4)
Changed in plainbox-provider-checkbox: | |
importance: | Undecided → High |
assignee: | nobody → Rod Smith (rodsmith) |
Changed in plainbox-provider-checkbox: | |
milestone: | none → 0.44.0 |
status: | New → Fix Committed |
Changed in plainbox-provider-checkbox: | |
status: | Fix Committed → Fix Released |