kvm: Fix macvtap naming issue. Closes #2023 #2181
Conversation
|
Review request. |
networking/kvm.go
Outdated
| mtu = config.MTU | ||
| } | ||
| nameTemplate := fmt.Sprintf("rkt-%s-vtap%%d", podID.String()[0:4]) | ||
| nameTemplate := fmt.Sprintf("rkt-%s-vtap%d", podID.String()[0:4], interface_number) |
There was a problem hiding this comment.
It looks like that it used to be some kind of a printf-template before (rkt-3729-vtap%d), but now it is not. Should the code expecting it as a template should be updated then (if there is such somewhere)? Also, the nameTemplate is misleading - it is not a template anymore, right?
There was a problem hiding this comment.
You are right - the idea was to push template to kernel side (what is working with other network drivers, but not with macvtap) and was modified to prepare this name directly on rkt side.
So, I should update variable name.
|
Reviewed, but it's not binding - not my area. |
e17fb34 to
5d4a4fb
Compare
|
Review request. |
| return errwrap.Wrap(errors.New("failed to enable forwarding"), err) | ||
| } | ||
|
|
||
| // patch plugin type only for single IPAM run time, then revert this change |
There was a problem hiding this comment.
With this method I can run underlying IPAM plugin to obtain IP configuration.
Without reverting this value after process, we loosed information about original plugin (saved later in .json) and this was a reason why gc without that does not work correctly in this configuration.
|
@steveej do you have time for review? |
|
@steveej ping? |
|
I am not sure that patching |
|
I'll do this tomorrow. |
|
rebase? |
5d4a4fb to
ce059a7
Compare
|
Done. |
|
Thanks! |
With @mstachowski help - this PR provides resolution to bug reported in #2023
We are still working on our testsuite (this work should turn into mergable PR due this or next week), which also covers this area.