stub setup for new ekman cluster
This commit is contained in:
Jonas Juselius
343
clusters/ekman/default.nix
Normal file
343
clusters/ekman/default.nix
Normal file
@@ -0,0 +1,343 @@
|
||||
let
|
||||
# Pin the deployment package-set to a specific version of nixpkgs
|
||||
# pkgs = import (builtins.fetchTarball {
|
||||
# url = "https://github.com/NixOS/nixpkgs/archive/e6377ff35544226392b49fa2cf05590f9f0c4b43.tar.gz";
|
||||
# sha256 = "1fra9wwy5gvj5ibayqkzqpwdf715bggc0qbmrfch4fghwvl5m70l";
|
||||
# }) {};
|
||||
pkgs = import <nixpkgs> {};
|
||||
|
||||
etcdNodes = {
|
||||
c0-0 = "10.255.240.200";
|
||||
c0-1 = "10.255.240.201";
|
||||
};
|
||||
|
||||
etcdCluster = {
|
||||
enable = true;
|
||||
existing = false;
|
||||
nodes = etcdNodes;
|
||||
};
|
||||
|
||||
nodes =
|
||||
with builtins;
|
||||
let nodes = genList (n: n + 1) 1; in
|
||||
map (n: ({ name = "c0-${toString n}"; address = "10.255.240.20${toString n}"; })) nodes;
|
||||
|
||||
ekman = {
|
||||
# deployment.tags = [ "frontend" ];
|
||||
node.myvnc = true;
|
||||
|
||||
systemd.targets = {
|
||||
sleep.enable = false;
|
||||
suspend.enable = false;
|
||||
hibernate.enable = false;
|
||||
hybrid-sleep.enable = false;
|
||||
};
|
||||
|
||||
features = {
|
||||
host = {
|
||||
address = "10.255.240.200";
|
||||
name = "c0-0";
|
||||
};
|
||||
|
||||
os = {
|
||||
externalInterface = "eno1";
|
||||
nfs.enable = true;
|
||||
nfs.exports = ''
|
||||
/exports 10.255.240.0/24(insecure,rw,sync,no_subtree_check,crossmnt,fsid=0,no_root_squash)
|
||||
/exports 10.255.241.0/24(insecure,rw,sync,no_subtree_check,crossmnt,fsid=0,no_root_squash)
|
||||
'';
|
||||
};
|
||||
|
||||
hpc = {
|
||||
slurm.server = true;
|
||||
frontend = true;
|
||||
};
|
||||
|
||||
k8s = {
|
||||
master.enable = true;
|
||||
node.enable = true;
|
||||
inherit nodes;
|
||||
inherit etcdCluster;
|
||||
};
|
||||
|
||||
monitoring = {
|
||||
server = {
|
||||
enable = false;
|
||||
scrapeHosts = [ "frontend" "bee0-0" ] ++ (builtins.map (x: x.name) nodes);
|
||||
defaultAlertReceiver = {
|
||||
email_configs = [
|
||||
{ to = "jonas.juselius@oceanbox.io"; }
|
||||
];
|
||||
};
|
||||
pageAlertReceiver = {
|
||||
webhook_configs = [
|
||||
{
|
||||
url = "https://prometheus-msteams.k2.itpartner.no/ekman";
|
||||
http_config = {
|
||||
tls_config = { insecure_skip_verify = true; };
|
||||
};
|
||||
}
|
||||
];
|
||||
};
|
||||
};
|
||||
webUI.enable = false;
|
||||
webUI.acmeEmail = "innovasjon@itpartner.no";
|
||||
webUI.allow = [
|
||||
"10.1.2.0/24"
|
||||
"172.19.254.0/24"
|
||||
"172.19.255.0/24"
|
||||
];
|
||||
infiniband-exporter = {
|
||||
enable = true;
|
||||
nameMap = ''
|
||||
0x0c42a10300ddc4bc "frontend"
|
||||
0x0c42a10300dbe7f4 "c0-1"
|
||||
'';
|
||||
};
|
||||
slurm-exporter = {
|
||||
enable = true;
|
||||
port = 6080;
|
||||
};
|
||||
};
|
||||
};
|
||||
|
||||
networking = {
|
||||
useDHCP = false;
|
||||
interfaces.enp33s0f0np0 = {
|
||||
useDHCP = false;
|
||||
ipv4.addresses = [ {
|
||||
address = "10.255.240.200";
|
||||
prefixLength = 24;
|
||||
} ];
|
||||
};
|
||||
# interfaces.enp33s0f0np1 = {
|
||||
# useDHCP = false;
|
||||
# ipv4.addresses = [ {
|
||||
# address = "10.1.61.100";
|
||||
# prefixLength = 24;
|
||||
# } ];
|
||||
# };
|
||||
# interfaces.ibp59s0 = {
|
||||
# useDHCP = false;
|
||||
# ipv4.addresses = [ {
|
||||
# address = "10.255.241.200";
|
||||
# prefixLength = 24;
|
||||
# } ];
|
||||
};
|
||||
defaultGateway = "10.255.240.1";
|
||||
firewall.extraCommands = ''
|
||||
iptables -I INPUT -s 10.255.241.0/24 -j ACCEPT
|
||||
iptables -t nat -A POSTROUTING -s 10.255.241.0/24 -j MASQUERADE
|
||||
'';
|
||||
};
|
||||
|
||||
fileSystems ={
|
||||
"/exports/home" = {
|
||||
device = "/home";
|
||||
options = [ "bind" ];
|
||||
};
|
||||
"/frontend" = {
|
||||
device = "/home";
|
||||
options = [ "bind" ];
|
||||
};
|
||||
# "/opt" = {
|
||||
# device = "10.255.63.80:/opt";
|
||||
# fsType = "nfs";
|
||||
# options = [ "soft" "rdma" "defaults" ];
|
||||
# };
|
||||
# "/data" = {
|
||||
# device = "10.255.63.80:/data";
|
||||
# fsType = "nfs";
|
||||
# options = [ "soft" "rdma" "defaults" ];
|
||||
# };
|
||||
# "/vol/local-storage/vol1" = {
|
||||
# device = "/vol/vol1";
|
||||
# options = [ "bind" ];
|
||||
# };
|
||||
# "/vol/local-storage/vol2" = {
|
||||
# device = "/vol/vol2";
|
||||
# options = [ "bind" ];
|
||||
# };
|
||||
};
|
||||
|
||||
nix.extraOptions = ''
|
||||
secret-key-files = /etc/nix/ekman.private
|
||||
'';
|
||||
|
||||
services.xserver = {
|
||||
enable = true;
|
||||
enableCtrlAltBackspace = true;
|
||||
layout = "us";
|
||||
xkbVariant = "altgr-intl";
|
||||
xkbOptions = "eurosign:e";
|
||||
displayManager = {
|
||||
gdm.enable = true;
|
||||
job.logToFile = true;
|
||||
};
|
||||
desktopManager.xfce.enable = true;
|
||||
};
|
||||
|
||||
services.prometheus.alertmanager.configuration.global = {
|
||||
smtp_smarthost = "smtpgw.itpartner.no:465";
|
||||
smtp_auth_username = "utvikling";
|
||||
smtp_auth_password = "S0m3rp0m@de#21!";
|
||||
smtp_hello = "ekman.oceanbox.io";
|
||||
smtp_from = "noreply@ekman.oceanbox.io";
|
||||
};
|
||||
|
||||
# services.nginx = {
|
||||
# virtualHosts = {
|
||||
# "ds.matnoc.regnekraft.io" = {
|
||||
# forceSSL = true;
|
||||
# enableACME = true;
|
||||
# serverAliases = [];
|
||||
# locations."/" = {
|
||||
# proxyPass = "http://localhost:9088";
|
||||
# proxyWebsockets = false;
|
||||
# extraConfig = ''
|
||||
# allow 10.1.2.0/24;
|
||||
# allow 172.19.254.0/24;
|
||||
# allow 172.19.255.0/24;
|
||||
# deny all;
|
||||
# '';
|
||||
# };
|
||||
# };
|
||||
# };
|
||||
# };
|
||||
|
||||
# services.gitlab-runner = {
|
||||
# enable = true;
|
||||
# extraPackages = with pkgs; [
|
||||
# singularity
|
||||
# ];
|
||||
# concurrent = 4;
|
||||
# services = {
|
||||
# sif = {
|
||||
# registrationConfigFile = "/var/lib/secrets/gitlab-runner-registration";
|
||||
# executor = "shell";
|
||||
# tagList = [ "ekman" "sif" ];
|
||||
# };
|
||||
# };
|
||||
# };
|
||||
|
||||
# security.sudo.extraConfig = ''
|
||||
# gitlab-runner ALL=(ALL) NOPASSWD: /run/current-system/sw/bin/singularity
|
||||
# '';
|
||||
|
||||
security.pam = {
|
||||
services.sshd.googleAuthenticator.enable = true;
|
||||
loginLimits = [
|
||||
{
|
||||
domain = "@users";
|
||||
item = "rss";
|
||||
type = "hard";
|
||||
value = 16000000;
|
||||
}
|
||||
{
|
||||
domain = "@users";
|
||||
item = "cpu";
|
||||
type = "hard";
|
||||
value = 180;
|
||||
}
|
||||
];
|
||||
};
|
||||
|
||||
# ssh-rsa is deprecated, but putty/winscp users use it
|
||||
# services.openssh.extraConfig = ''
|
||||
# pubkeyacceptedalgorithms ssh-rsa,ssh-ed25519-cert-v01@openssh.com,ecdsa-sha2-nistp256-cert-v01@openssh.com,ecdsa-sha2-nistp384-cert-v01@openssh.com,ecdsa-sha2-nistp521-cert-v01@openssh.com,sk-ssh-ed25519-cert-v01@openssh.com,sk-ecdsa-sha2-nistp256-cert-v01@openssh.com,rsa-sha2-512-cert-v01@openssh.com,rsa-sha2-256-cert-v01@openssh.com,ssh-ed25519,ecdsa-sha2-nistp256,ecdsa-sha2-nistp384,ecdsa-sha2-nistp521,sk-ssh-ed25519@openssh.com,sk-ecdsa-sha2-nistp256@openssh.com,rsa-sha2-512,rsa-sha2-256
|
||||
# '';
|
||||
|
||||
imports = [ ./cluster.nix ./hw/frontend.nix ];
|
||||
};
|
||||
|
||||
compute = {
|
||||
# deployment.tags = [ "compute" ];
|
||||
|
||||
fileSystems = {
|
||||
"/frontend" = {
|
||||
device = "10.255.240.200:/home";
|
||||
fsType = "nfs";
|
||||
options = [
|
||||
"soft"
|
||||
"defaults"
|
||||
"noauto"
|
||||
"x-systemd.automount"
|
||||
];
|
||||
};
|
||||
# "/opt" = {
|
||||
# device = "10.1.63.80:/opt";
|
||||
# fsType = "nfs";
|
||||
# options = [ "soft" "rdma" "defaults" ];
|
||||
# };
|
||||
# "/data" = {
|
||||
# device = "10.1.63.80:/data";
|
||||
# fsType = "nfs";
|
||||
# options = [ "soft" "rdma" "defaults" ];
|
||||
# };
|
||||
};
|
||||
|
||||
systemd.automounts = [
|
||||
{
|
||||
where = "/frontend";
|
||||
wantedBy = [ "default.target" ];
|
||||
}
|
||||
];
|
||||
};
|
||||
|
||||
mkCompute = host:
|
||||
let
|
||||
ipoib = builtins.replaceStrings [".240."] [".241."] host.address;
|
||||
hw = ./hw + "/${host.name}.nix";
|
||||
in {
|
||||
"${host.name}" = {
|
||||
features = {
|
||||
inherit host;
|
||||
os.externalInterface = "enp33s0f0np0";
|
||||
hpc.compute = true;
|
||||
k8s = { inherit etcdCluster; };
|
||||
};
|
||||
|
||||
node = {
|
||||
i40efix = true;
|
||||
};
|
||||
|
||||
networking = {
|
||||
useDHCP = false;
|
||||
interfaces.enp33s0f0np0 = {
|
||||
useDHCP = false;
|
||||
ipv4.addresses = [ {
|
||||
address = host.address;
|
||||
prefixLength = 24;
|
||||
} ];
|
||||
# ipv4.routes = [ {
|
||||
# address = "10.1.62.2";
|
||||
# prefixLength = 32;
|
||||
# via = "10.1.61.100";
|
||||
# } ];
|
||||
|
||||
};
|
||||
# interfaces.ibp65s0 = {
|
||||
# useDHCP = false;
|
||||
# ipv4.addresses = [ {
|
||||
# address = ipoib;
|
||||
# prefixLength = 24;
|
||||
# } ];
|
||||
# };
|
||||
};
|
||||
imports = [ ./cluster.nix hw ];
|
||||
}
|
||||
// compute;
|
||||
};
|
||||
in {
|
||||
## morph
|
||||
# network = {
|
||||
# inherit pkgs;
|
||||
# description = "ekman";
|
||||
# ordering = {
|
||||
# tags = [ "frontend" "compute" ];
|
||||
# };
|
||||
# };
|
||||
|
||||
inherit ekman;
|
||||
} // builtins.foldl' (a: n: a // mkCompute n) {} nodes
|
||||
|
||||
Reference in New Issue
Block a user