Linkerd Joins the Cloud Native Computing Foundation | Hacker News Reader

Linkerd Joins the Cloud Native Computing Foundation

blog.buoyant.io

80 pointslew2112 comments

The killer application of linkerd is kubernetes, where a lot of people are using linkerd to get around using Kubernetes "Services". I think the Ingress controllers talk directly to Linkerd.

I have requested for L4 support in Linkerd that will finally make it possible for Linkerd to actually become an ingress + Service abstraction.

Op, mind telling us why people are using linkerd to get around using Kubernetes "Services" ?

there's quite a bit here - https://monzo.com/blog/2016/09/19/building-a-modern-bank-bac...

and

https://blog.buoyant.io/2016/10/14/a-service-mesh-for-kubern...

But you should ping linkerd on twitter. They are really responsive.

Not sure about this one yet - some benefits, but not really massive traction. I don’t clearly see the benefits (although latency-sensitive internal loadbalancing seems quite neat), but it’s yet-another proxy in the neverending proxying of proxies within K8S already.

Hi, from what I read the linkerd proxy allow to supervise response stats (latency, error rate). In kubernetes you didn't have global proxy stats. It's a circuit breaker and you can connect with other service with different protocols (http,rpc) without have to know what protocol other service use. I didn't have tested linkerd yet but I noticed this points. I think it's really useful when you have micro services

" it’s yet-another proxy in the neverending proxying of proxies within K8S already." - <3

I think it'd be nice to see the functionality of linkerd merged into k8s.

To be fair, most proxies are in-kernel in the form of *NAT.

I use ebay/Fabio in a similar pattern as Linkerd and it's been amazing. I can understand that it's very useful to also have the latency stats and circuit breaker features like netflix/Hystrix.

There are also more and more private networking / tunnel / sdn tools popping up, I wonder which of the techniques will win out. There are advantages to not doing too much in each tool.

It's useful to separate traffic to instances/services by need-to-know e.g. admins vs users, and that would require two instances of Linkerd to truly separate the traffic. In this case I use VLANs/SDNs with service discovery instead.

Always interested to hear how others are solving this routing problem.

Howdy, Linkerd person here. Always nice to have a surprise Hacker News frontpage post! Please let me know if you have any questions. We also have a very active Slack at slack.linkerd.io. Hop on in and say hello.

We're using Linkerd and Helium along with DCOS/Mesos at ForeSee - We're super happy with the technology and we look forward seeing Linkerd supporting more protocols.

We are using Linkerd at NextVR and we are where happy it's L5 routing and other features. Soon are looking forward to integrate Zipkin with GCE tracer.

Reply on news.ycombinator.com