yahoo-eng-team team mailing list archive

[Ilya Chukhnakov <1930200@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

This RFE is to add a new type of IP resource to OpenStack Neutron - a
virtual IP that can be shared across multiple ports/VMs (similar to
anycast IP) and is guaranteed to only be reachable within the same
physical server/node boundaries.

E.g. assuming the cloud has the following resources:
- node "Node-A" with 2 virtual machines with corresponding ports "Port-A-Src" and "Port-A-Dest"
- node "Node-B" with 2 virtual machines with corresponding ports "Port-B-Src" and "Port-B-Dest"
- node "Node-C" with 1 virtual machine with port "Port-C-Src"
We would like to be able to create an IP address resource "NodeLocal-IP" and bind it to "Port-A-Dest" and "Port-B-Dest" such that:
- packets originating from "Port-A-Src" will be guaranteed to be only sent to "Port-A-Dest"
- packets originating from "Port-B-Src" will be guaranteed to be only sent to "Port-B-Dest"
- packets originating from "Port-C-Src" will not reach any destination service

Some examples/use-cases for the feature:
- (generic) Side-car proxies, load-balancers or service mesh platforms. For example K8s Istio project runs a side-car Envoy proxies on each node and all packets from client containers are transparently re-routed through those local proxies. If the proposed feature is added to OpenStack Neutron, that could enable and/or simplify the integration and development of such services for OpenStack.
- (large-scale/high-performance) Distributed caches and CDNs. If the proposed feature is added, that could enable or greatly simplify the use of distributed caches on OpenStack clouds, for example running a VM with Docker Registry cache (passthrough mode) on each physical node could greatly speedup the download/startup time for containers running inside OpenStack VMs.
- (OpenStack) The feature can potentially benefit the OpenStack Octavia project to improve the scalability for very-large-scale deployments by applying load-balance-at-source pattern.

Co-Authored-By: Oleg Bondarev

** Affects: neutron
     Importance: Undecided
         Status: New


** Tags: rfe

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1930200

Title:
  [RFE] Add support for Node-Local virtual IP

Status in neutron:
  New

Bug description:
  This RFE is to add a new type of IP resource to OpenStack Neutron - a
  virtual IP that can be shared across multiple ports/VMs (similar to
  anycast IP) and is guaranteed to only be reachable within the same
  physical server/node boundaries.

  E.g. assuming the cloud has the following resources:
  - node "Node-A" with 2 virtual machines with corresponding ports "Port-A-Src" and "Port-A-Dest"
  - node "Node-B" with 2 virtual machines with corresponding ports "Port-B-Src" and "Port-B-Dest"
  - node "Node-C" with 1 virtual machine with port "Port-C-Src"
  We would like to be able to create an IP address resource "NodeLocal-IP" and bind it to "Port-A-Dest" and "Port-B-Dest" such that:
  - packets originating from "Port-A-Src" will be guaranteed to be only sent to "Port-A-Dest"
  - packets originating from "Port-B-Src" will be guaranteed to be only sent to "Port-B-Dest"
  - packets originating from "Port-C-Src" will not reach any destination service

  Some examples/use-cases for the feature:
  - (generic) Side-car proxies, load-balancers or service mesh platforms. For example K8s Istio project runs a side-car Envoy proxies on each node and all packets from client containers are transparently re-routed through those local proxies. If the proposed feature is added to OpenStack Neutron, that could enable and/or simplify the integration and development of such services for OpenStack.
  - (large-scale/high-performance) Distributed caches and CDNs. If the proposed feature is added, that could enable or greatly simplify the use of distributed caches on OpenStack clouds, for example running a VM with Docker Registry cache (passthrough mode) on each physical node could greatly speedup the download/startup time for containers running inside OpenStack VMs.
  - (OpenStack) The feature can potentially benefit the OpenStack Octavia project to improve the scalability for very-large-scale deployments by applying load-balance-at-source pattern.

  Co-Authored-By: Oleg Bondarev

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1930200/+subscriptions