CadolesKube/typhoon
7
0
Fork 0
mirror of https://github.com/puppetmaster/typhoon.git synced 2025-08-11 08:46:04 +02:00
Code Issues Packages Projects Releases Wiki Activity

Combine NLBs to use one NLB per cluster

Browse Source 
* Simplify clusters to come with a single NLB
* Listen for apiserver traffic on port 6443 and forward
to controllers (with healthy apiserver)
* Listen for ingress traffic on ports 80/443 and forward
to workers (with healthy ingress controller)
* Reduce cost of default clusters by 1 NLB ($18.14/month)
* Keep using CNAME records to the `ingress_dns_name` NLB and
the nginx-ingress addon for Ingress (up to a few million RPS)
* Users with heavy traffic (many million RPS) can create their
own separate NLB(s) for Ingress and use the new output worker
target groups
* Fix issue where additional worker pools come with an
extraneous network load balancer
This commit is contained in:
Dalton Hubble
2018-06-18 22:56:32 -07:00
parent f4d3059b00
commit 316f06df06
9 changed files with 113 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
CHANGES.md
View File

@ -10,7 +10,13 @@ Notable changes between versions.
#### AWS
* Switch `kube-apiserver` port from 443 to 6443 ([#248](https://github.com/poseidon/typhoon/pull/248))
* Update NLB, security groups, and generated kubeconfig's
* Combine apiserver and ingress NLBs ([#249](https://github.com/poseidon/typhoon/pull/249))
* Simplify clusters to come with one NLB. Reduce cost by ~$18/month per cluster.
* Users may keep using CNAME records to `ingress_dns_name` and the `nginx-ingress` addon for Ingress (up to a few million RPS)
* Users with heavy traffic (many million RPS) should create a separate NLB(s) for Ingress instead
* Listen for apiserver traffic on port 6443 and forward to controllers (with healthy apiserver)
* Listen for ingress traffic on ports 80/443 and forward to workers (with healthy ingress controller)
* Worker pools (advanced) no longer include an extraneous load balancer
#### Bare-Metal