Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Enable GLB Healthcheck to announce VIP/Bind status into BGP #92

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

robloxrob
Copy link

This pull request adds a new feature to GLB healthcheck. It allows it to locally announce the status of a VIP/bind into BGP using a local instance of GoBGP. It integrates with a locally deployed GoBGP daemon to announce the VIP/Bind into BGP when at least one backend is marked healthy. This allows you to prevent a statically announced route via BIRD to black hole traffic for a GLB instance that is unable to talk to its backends. In the event that all backends are marked either inactive or marked as unhealthy GLB healthcheck will withdraw the route. The feature is configurable (see included documentation) and defaults to off. We have been running this in production for a while and it simplified our integration with BGP. We have been running this for a while in prod, but I have been busy and been unable to issue a pull request.

I needed to change how some of the testing works to get the full suite to run correctly.

  • In Vagrantfile I named all of the shell steps to simplify debugging
  • Moved building GLB healthcheck to Golang 1.11 to allow compilation to work for the newer included libraries
  • Because Debian has a new kernel and the older kernel headers weren't available I correct this and reload the VMs to ensure that the changes are picked up

Worked with @ravisinghsfbay on correcting parts of the testing. We both work in the same team here at Roblox.

This pull request adds a new feature to GLB healthcheck. It allows it to locally announce the status of a VIP/bind into BGP using a local instance of GoBGP. It integrates with a locally deployed GoBGP daemon to announce the VIP/Bind into BGP when at least one backend is marked healthy. This allows you to prevent a statically announced route via BIRD to black hole traffic for a GLB instance that is unable to talk to its backends. In the event that all backends are marked either inactive or marked as unhealthy GLB healthcheck will withdraw the route. The feature is configurable (see included documentation) and defaults to off. We have been running this in production for a while and it simplified our integration with BGP.
Merge Golang fixes
@mcarbonneaux
Copy link

it would be very practical. this allows better exploitation and faster bgb convergence.

@Slepwin
Copy link

Slepwin commented Nov 10, 2022

Hi GitHub folks, could you please merge thisPR.

@newerdog
Copy link

newerdog commented Dec 8, 2022

Hi GitHub folks, could you please merge thisPR.

Withdrawing the route is the best practice used in commercial LB like F5\Netscaller, but also in Google SeeSawV2 project. However, I prefer the method with redistribute kerner routes. It is native way without any config changing

In GObgp is
redistribute-route-type-list = ["connect"]

Also i`d recommend setting up BFD to remove routes in a tick when Director goes Down

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants