mii-mon should have a consistent unit schema
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Netplan |
Fix Released
|
High
|
Unassigned | ||
netplan.io (Ubuntu) |
Fix Released
|
High
|
Unassigned | ||
Xenial |
Invalid
|
Undecided
|
Unassigned | ||
Artful |
Invalid
|
Undecided
|
Unassigned | ||
Bionic |
Fix Released
|
Undecided
|
Unassigned | ||
nplan (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Xenial |
Fix Released
|
Undecided
|
Unassigned | ||
Artful |
Fix Released
|
Undecided
|
Unassigned | ||
Bionic |
Invalid
|
Undecided
|
Unassigned |
Bug Description
[Impact]
Documentation is confusing to users wishing to tune bond and bridge parameters. This potentially affects all users of netplan.
[Test case]
1) Update the netplan.io package.
2) Run 'man netplan'
3) Verify that the documentation under "parameters" for bridges and bonds is clear that time-based parameters are in milliseconds unless otherwise specified by the individual parameter:
For example, in the bonds section:
- learn-packet-
- mii-monitor-
[Regression potential]
Change is limited to the manpage; there is no risk of regression there, except for a garbled manpage or missing file. Any new bugs in netplan behavior found after this SRU that can't be reproduced with the immediately previous version should be investigated as potential toolchain / build issues introduced by the new build required for the SRU.
---
The docs say "Using the NetworkManager renderer, parameter values for intervals should be expressed in milliseconds; for the systemd renderer, they should be in seconds unless otherwise specified."
This however fails to encapsulate the description of the network in a form abstracted from the renderer. We should have a scheme which is clear, has one and only one explicit form, and works with all renderers.
Changed in netplan: | |
status: | New → Triaged |
importance: | Undecided → High |
description: | updated |
The manpage is missing a doc update here, clearly.
We made sure the units were consistently milliseconds where appropriate (some parameters are always seconds, no matter what backend is used) late in the bionic cycle (LP: #1765833); the last of these updates included a doc change, but it didn't change it by much.
Let's fix this in a SRU.