Allow Calico networking on Azure and DigitalOcean

* Introduce "calico" as a `networking` option on Azure and DigitalOcean
using Calico's new VXLAN support (similar to flannel). Flannel remains
the default on these platforms for now.
* Historically, DigitalOcean and Azure only allowed Flannel as the
CNI provider, since those platforms don't support IPIP traffic that
was previously required for Calico.
* Looking forward, its desireable for Calico to become the default
across Typhoon clusters, since it provides NetworkPolicy and a
consistent experience
* No changes to AWS, GCP, or bare-metal where Calico remains the
default CNI provider. On these platforms, IPIP mode will always
be used, since its available and more performant than vxlan
This commit is contained in:
Dalton Hubble 2019-05-06 00:38:23 -07:00
parent b9bab739ce
commit 147c21a4bd
10 changed files with 56 additions and 20 deletions

View File

@ -7,14 +7,23 @@ Notable changes between versions.
* Kubernetes [v1.14.2](https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG-1.14.md#v1142)
* Update etcd from v3.3.12 to [v3.3.13](https://github.com/etcd-io/etcd/releases/tag/v3.3.13)
* Upgrade Calico from v3.6.1 to [v3.7.2](https://docs.projectcalico.org/v3.7/release-notes/)
* Change flannel port from 8472 (kernel default) to 4789 (IANA VXLAN)
* Change flannel VXLAN port from 8472 (kernel default) to 4789 (IANA VXLAN)
#### AWS
* Only set internal VXLAN rules when `networking` is flannel (default: calico)
* Only set internal VXLAN rules when `networking` is "flannel" (default: calico)
#### Azure
* Allow choosing Calico as the network provider (experimental) ([#472](https://github.com/poseidon/typhoon/pull/472))
* Add a `networking` variable accepting "flannel" (default) or "calico"
* Use VXLAN encapsulation since Azure doesn't support IPIP
#### DigitalOcean
* Allow choosing Calico as the network provider (experimental) ([#472](https://github.com/poseidon/typhoon/pull/472))
* Add a `networking` variable accepting "flannel" (default) or "calico"
* Use VXLAN encapsulation since DigitalOcean doesn't support IPIP
* Add explicit ordering between firewall rule creation and secure copying Kubelet credentials ([#469](https://github.com/poseidon/typhoon/pull/469))
* Fix race scenario if copies to nodes were before rule creation, blocking cluster creation

View File

@ -2,11 +2,18 @@
module "bootkube" {
source = "git::https://github.com/poseidon/terraform-render-bootkube.git?ref=85571f6dae3522e2a7de01b7e0a3f7e3a9359641/"
cluster_name = "${var.cluster_name}"
api_servers = ["${format("%s.%s", var.cluster_name, var.dns_zone)}"]
etcd_servers = ["${formatlist("%s.%s", azurerm_dns_a_record.etcds.*.name, var.dns_zone)}"]
asset_dir = "${var.asset_dir}"
networking = "flannel"
cluster_name = "${var.cluster_name}"
api_servers = ["${format("%s.%s", var.cluster_name, var.dns_zone)}"]
etcd_servers = ["${formatlist("%s.%s", azurerm_dns_a_record.etcds.*.name, var.dns_zone)}"]
asset_dir = "${var.asset_dir}"
networking = "${var.networking}"
# only effective with Calico networking
# we should be able to use 1450 MTU, but in practice, 1410 was needed
network_encapsulation = "vxlan"
network_mtu = "1410"
pod_cidr = "${var.pod_cidr}"
service_cidr = "${var.service_cidr}"
cluster_domain_suffix = "${var.cluster_domain_suffix}"

View File

@ -88,6 +88,12 @@ variable "asset_dir" {
type = "string"
}
variable "networking" {
description = "Choice of networking provider (flannel or calico)"
type = "string"
default = "flannel"
}
variable "host_cidr" {
description = "CIDR IPv4 range to assign to instances"
type = "string"

View File

@ -2,12 +2,17 @@
module "bootkube" {
source = "git::https://github.com/poseidon/terraform-render-bootkube.git?ref=85571f6dae3522e2a7de01b7e0a3f7e3a9359641/"
cluster_name = "${var.cluster_name}"
api_servers = ["${format("%s.%s", var.cluster_name, var.dns_zone)}"]
etcd_servers = "${digitalocean_record.etcds.*.fqdn}"
asset_dir = "${var.asset_dir}"
networking = "flannel"
network_mtu = 1440
cluster_name = "${var.cluster_name}"
api_servers = ["${format("%s.%s", var.cluster_name, var.dns_zone)}"]
etcd_servers = "${digitalocean_record.etcds.*.fqdn}"
asset_dir = "${var.asset_dir}"
networking = "${var.networking}"
# only effective with Calico networking
network_encapsulation = "vxlan"
network_mtu = "1450"
pod_cidr = "${var.pod_cidr}"
service_cidr = "${var.service_cidr}"
cluster_domain_suffix = "${var.cluster_domain_suffix}"

View File

@ -1,6 +1,7 @@
# Secure copy etcd TLS assets and kubeconfig to controllers. Activates kubelet.service
resource "null_resource" "copy-controller-secrets" {
count = "${var.controller_count}"
depends_on = [
"digitalocean_firewall.rules",
]

View File

@ -71,6 +71,12 @@ variable "asset_dir" {
type = "string"
}
variable "networking" {
description = "Choice of networking provider (flannel or calico)"
type = "string"
default = "flannel"
}
variable "pod_cidr" {
description = "CIDR IPv4 range to assign Kubernetes pods"
type = "string"

View File

@ -1,6 +1,7 @@
# Secure copy etcd TLS assets and kubeconfig to controllers. Activates kubelet.service
resource "null_resource" "copy-controller-secrets" {
count = "${var.controller_count}"
depends_on = [
"digitalocean_firewall.rules",
]

View File

@ -253,6 +253,7 @@ Reference the DNS zone with `"${azurerm_dns_zone.clusters.name}"` and its resour
| worker_priority | Set priority to Low to use reduced cost surplus capacity, with the tradeoff that instances can be deallocated at any time | Regular | Low |
| controller_clc_snippets | Controller Container Linux Config snippets | [] | [example](/advanced/customization/#usage) |
| worker_clc_snippets | Worker Container Linux Config snippets | [] | [example](/advanced/customization/#usage) |
| networking | Choice of networking provider | "flannel" | "flannel" or "calico" (experimental) |
| host_cidr | CIDR IPv4 range to assign to instances | "10.0.0.0/16" | "10.0.0.0/20" |
| pod_cidr | CIDR IPv4 range to assign to Kubernetes pods | "10.2.0.0/16" | "10.22.0.0/16" |
| service_cidr | CIDR IPv4 range to assign to Kubernetes services | "10.3.0.0/16" | "10.3.0.0/24" |

View File

@ -253,6 +253,7 @@ Digital Ocean requires the SSH public key be uploaded to your account, so you ma
| image | Container Linux image for instances | "coreos-stable" | coreos-stable, coreos-beta, coreos-alpha |
| controller_clc_snippets | Controller Container Linux Config snippets | [] | [example](/advanced/customization/) |
| worker_clc_snippets | Worker Container Linux Config snippets | [] | [example](/advanced/customization/) |
| networking | Choice of networking provider | "flannel" | "flannel" or "calico" (experimental) |
| pod_cidr | CIDR IPv4 range to assign to Kubernetes pods | "10.2.0.0/16" | "10.22.0.0/16" |
| service_cidr | CIDR IPv4 range to assign to Kubernetes services | "10.3.0.0/16" | "10.3.0.0/24" |
| cluster_domain_suffix | FQDN suffix for Kubernetes services answered by coredns. | "cluster.local" | "k8s.example.com" |

View File

@ -26,20 +26,19 @@ Network performance varies based on the platform and CNI plugin. `iperf` was use
|----------------------------|-------:|-------------:|-------------:|
| AWS (flannel) | 5 Gb/s | 4.94 Gb/s | 4.89 Gb/s |
| AWS (calico, MTU 1480) | 5 Gb/s | 4.94 Gb/s | 4.42 Gb/s |
| AWS (calico, MTU 8981) | 5 Gb/s | 4.94 Gb/s | 4.75 Gb/s |
| Azure (flannel) | Varies | 749 Mb/s | 680 Mb/s |
| AWS (calico, MTU 8981) | 5 Gb/s | 4.94 Gb/s | 4.90 Gb/s |
| Azure (flannel) | Varies | 749 Mb/s | 650 Mb/s |
| Azure (calico) | Varies | 749 Mb/s | 650 Mb/s |
| Bare-Metal (flannel) | 1 Gb/s | 940 Mb/s | 903 Mb/s |
| Bare-Metal (calico) | 1 Gb/s | 940 Mb/s | 931 Mb/s |
| Bare-Metal (flannel, bond) | 3 Gb/s | 2.3 Gb/s | 1.17 Gb/s |
| Bare-Metal (calico, bond) | 3 Gb/s | 2.3 Gb/s | 1.17 Gb/s |
| Digital Ocean | 2 Gb/s | 1.97 Gb/s | 1.64 Gb/s |
| Digital Ocean (flannel) | Varies | 1.97 Gb/s | 1.20 Gb/s |
| Digital Ocean (calico) | Varies | 1.97 Gb/s | 1.20 Gb/s |
| Google Cloud (flannel) | 2 Gb/s | 1.94 Gb/s | 1.76 Gb/s |
| Google Cloud (calico) | 2 Gb/s | 1.94 Gb/s | 1.81 Gb/s |
Notes:
* Calico and Flannel have comparable performance. Platform and configuration differences dominate.
* AWS and Azure node bandwidth (i.e. upper bound) depends greatly on machine type
* Azure and DigitalOcean network performance can be quite variable or depend on machine type
* Only [certain AWS EC2 instance types](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/network_mtu.html#jumbo_frame_instances) allow jumbo frames. This is why the default MTU on AWS must be 1480.
* Neither CNI provider seems to be able to leverage bonded NICs well (bare-metal)