Update proxycfg state management and xDS generation for transparent proxy #9894
Conversation
By accepting a name the function can be used for other inbound listeners, like the one for TransparentProxy.
| } | ||
|
|
||
| if chain == nil || chain.IsDefault() { | ||
| // TODO(rb): make this do the old school stuff too |
There was a problem hiding this comment.
What old stuff was this?
| if u.CentrallyConfigured { | ||
| continue | ||
| } | ||
| snap.ConnectProxy.UpstreamConfig[u.Identifier()] = &u |
There was a problem hiding this comment.
Note you are taking the address of the loop variable here, not the spot in the slice.
There was a problem hiding this comment.
I think now you're taking the address of a copy of the slice item.
I think the fix would be = &s.proxyCfg.Upstreams[i]
| @@ -54,35 +58,225 @@ func (s *Server) listenersFromSnapshot(cInfo connectionInfo, cfgSnap *proxycfg.C | |||
|
|
|||
| // listenersFromSnapshotConnectProxy returns the "listeners" for a connect proxy service | |||
| func (s *Server) listenersFromSnapshotConnectProxy(cInfo connectionInfo, cfgSnap *proxycfg.ConfigSnapshot) ([]proto.Message, error) { | |||
| // One listener for each upstream plus the public one | |||
| resources := make([]proto.Message, len(cfgSnap.Proxy.Upstreams)+1) | |||
| resources := make([]proto.Message, 1) | |||
There was a problem hiding this comment.
Is there a better guesstimate that we can use here now?
There was a problem hiding this comment.
There isn't a great one. If in TProxy mode there might only be 2 listeners, since it relies on per-upstream filter chains rather than per-upstream listeners. Outside of TProxy mode it could be the size of the DiscoverChain map.
Not sure if it's worth doing having a variable starting capacity based on whether TProxy is on or not.
| cfgSnap *proxycfg.ConfigSnapshot, | ||
| tlsContext *envoy_tls_v3.DownstreamTlsContext, | ||
| ) (*envoy_listener_v3.FilterChain, error) { | ||
| // TODO (freddy) Make this actually legible |
There was a problem hiding this comment.
I feel like there's some work we could do during ConfigSnapshot construction to alleviate some of this pain down here in the xDS code.
|
|
||
| // Only create the outbound listener when there are upstreams and filter chains are present | ||
| if outboundListener != nil && hasChains { | ||
| // Filter chains are stable sorted to avoid draining if the list is provided out of order |
|
🍒 If backport labels were added before merging, cherry-picking will start automatically. To retroactively trigger a backport after merging, add backport labels and re-run https://circleci.com/gh/hashicorp/consul/339575. |
This PR makes a few changes: