Check if we are on ipv4, ipv6 or dualStack when doing tailscale #7838
Conversation
manuelbuil
force-pushed
the
ipv4ipv6tailscale
branch
2 times, most recently
from
d9141bf
to
28de47b
Compare
manuelbuil
force-pushed
the
ipv4ipv6tailscale
branch
2 times, most recently
from
7849673
to
bbd49f6
Compare
manuelbuil
force-pushed
the
ipv4ipv6tailscale
branch
from
bbd49f6
to
3a239e1
Compare
Codecov ReportPatch coverage has no change and project coverage change:
Additional details and impacted files@@ Coverage Diff @@
## master #7838 +/- ##
==========================================
+ Coverage 47.18% 51.45% +4.27%
==========================================
Files 143 143
Lines 14509 14532 +23
==========================================
+ Hits 6846 7478 +632
+ Misses 6576 5865 -711
- Partials 1087 1189 +102
Flags with carried forward coverage won't be shown. Click here to find out more.
☔ View full report in Codecov by Sentry. |
pkg/util/net.go
Outdated
| @@ -97,10 +97,9 @@ func GetFirst6(elems []net.IP) (net.IP, error) { | |||
| // If no IPv6 addresses are found, an error is raised. | |||
| func GetFirst6Net(elems []*net.IPNet) (*net.IPNet, error) { | |||
| for _, elem := range elems { | |||
| if elem == nil || elem.IP.To16() == nil { | |||
| continue | |||
| if elem != nil || netutils.IsIPv6(elem.IP) { | |||
There was a problem hiding this comment.
This is wrong, you'll now return any non-nil element! The previous code was skipping non-ipv6 addresses; I'm not sure how this is functionally different?
| if elem != nil || netutils.IsIPv6(elem.IP) { | |
| if elem != nil && netutils.IsIPv6(elem.IP) { |
Also, note that the code you're calling effectively does the same thing - checks for non-nil .To16():
https://github.com/kubernetes/utils/blob/9f6742963106fb1fadd7ec2493d6d6cd689e5318/net/ipfamily.go#L106
There was a problem hiding this comment.
good catch! I missed that 🤦
It is slightly different in the ipfamily.go code because by the time to code reaches case ip.To16() != nil:, ipv4 ips were already returned in the previous line, i.e. only ipv6 addresses can reach case ip.To16() != nil:
There was a problem hiding this comment.
Run this gist and you'll see the difference: https://gist.github.com/manuelbuil/73d7e0ff32a69eb045602aeee45896a6
Signed-off-by: Manuel Buil <mbuil@suse.com>
manuelbuil
force-pushed
the
ipv4ipv6tailscale
branch
from
3a239e1
to
f21a014
Compare
Proposed Changes
Instead of listing the IPs of the tailscale interface in the variable
IPs []net.IP, separate the ipv4 and ipv6 address that are always assigned to the tailscale interface:That way we can easily control what address to use in kube-api depending on the configured mode in k3s.
Also a bug in utils/net.go was fixed:
is not a correct way to check if an IP is ipv6. It works for both ipv4 and ipv6!
Types of Changes
Bugfix
Verification
Verify that this works in:
Deploy k3s in three modes:
And check what IP is used for kubeapi-server in the flag
advertised-ip. It should be ipv4 for ipv4-only and dual-stack mode. Whereas it should be ipv6 for ipv6-only mode.Testing
Linked Issues
#7831
User-Facing Change
Further Comments