Support IPv6 VPC-private addresses for network interfaces #9269
Conversation
bnaecker
force-pushed
the
vpc-private-ipv6-address-queries
branch
2 times, most recently
from
4222a41 to
189254f
Compare
- Make more explicit how we assign IP addresses for NICs, by adding a full configuration type. That includes details of whether an address is auto assigned or explicitly requested, and the corresponding transit IPs. Also adds the notion of single-stack IPv4 or IPv6 NICs, or explicitly dual-stack. - Restructure some of the internals of the network interface `InsertQuery` to more clearly generate SQL based on the assignment preferences - Add `NextIpv6Address` type and use it internally to create VPC-private IPv6 addresses for interfaces. Add a few basic tests for the new behavior. - Rework callers of `IncompleteNetworkInterface` to use new IP configuration enum. - Closes #9243 - Closes #9244
bnaecker
force-pushed
the
vpc-private-ipv6-address-queries
branch
from
189254f to
d1f5170
Compare
bnaecker
changed the title
| ), | ||
| }, | ||
| None, //Request IP address assignment | ||
| IpConfig::auto_ipv4(), |
There was a problem hiding this comment.
Not for this PR but a general thought on this. I think what we'll want as a default is an IpConfig::auto_dual_stack()? Same for instances.
There was a problem hiding this comment.
Yeah, that sounds reasonable to me. There's no real reason to prevent VPC-private IPv6 addressing, since the address space is so huge.
| InsertError::IpAddressNotAvailable(ip) | ||
| let Some(ipv4) = interface.ip_config.ipv4_addr() else { | ||
| let err = Error::internal_error(&format!( | ||
| "Violated constraint that ensures unique \ |
There was a problem hiding this comment.
Does this mean auto assignment failed?
There was a problem hiding this comment.
Not quite, or at least that's not what I was going for. This means someone tried to insert an explicit IPv4 address, we filtered it out because it's already been used, and so we tried to insert NULL for both the IPv4 and IPv6 addresses.
So it's really "you requested a specific address, but it was already used".
There was a problem hiding this comment.
To put a finer point on it, this could be an assert. We made a programming error somewhere.
- IP-version specific first / last address functions - Fixup some names and typos - Allow transit IPs on an instance NIC - Add default dual-stack constructor
2 participants
InsertQueryto more clearly generate SQL based on the assignment preferencesNextIpv6Addresstype (needs tests)IncompleteNetworkInterfaceto use new assignment-request enum