unified cluster setup using colmena

2022-09-09 11:00:55 +02:00
parent ff19c52fd1
commit 980364adee
61 changed files with 966 additions and 117 deletions
--- a/cluster/nfs1/default.nix
+++ b/cluster/nfs1/default.nix
@@ -0,0 +1,152 @@
+{ pkgs ? import <nixpkgs> {} }:
+let
+  # Pin the deployment package-set to a specific version of nixpkgs
+  # pkgs = import (builtins.fetchTarball {
+  #   url = "https://github.com/NixOS/nixpkgs/archive/e9148dc1c30e02aae80cc52f68ceb37b772066f3.tar.gz";
+  #   sha256 = "1ckzhh24mgz6jd1xhfgx0i9mijk6xjqxwsshnvq789xsavrmsc36";
+  # }) {};
+  # pkgs = import <nixpkgs> {};
+  etcdCluster = import ../etcdCluster.nix;
+  name = "nfs1";
+  address = "10.255.241.90";
+in {
+  nfs1 = { config, pkgs, ... }: with pkgs; {
+    deployment.tags = [ "fs" "fs-work" ];
+    deployment.targetHost = address;
+    system.autoUpgrade.enable = lib.mkForce false;
+
+    systemd.targets = {
+      sleep.enable = false;
+      suspend.enable = false;
+      hibernate.enable = false;
+      hybrid-sleep.enable = false;
+    };
+
+    services.udev.extraRules = ''
+        KERNEL=="ibp65s0", SUBSYSTEM=="net", ATTR{create_child}:="0x7666"
+    '';
+
+    environment.systemPackages = with pkgs; [
+        rdma-core
+        hwloc
+    ];
+
+    cluster.k8sNode = true;
+
+    features = {
+      host = {
+        inherit address;
+        inherit name;
+      };
+
+      os = {
+        externalInterface = "enp33s0f3np3";
+        nfs.enable = true;
+        nfs.exports = ''
+          /exports 10.255.241.0/24(insecure,rw,async,no_subtree_check,crossmnt,fsid=0,no_root_squash)
+          /exports 10.255.243.0/24(insecure,rw,async,no_subtree_check,crossmnt,fsid=0,no_root_squash)
+        '';
+      };
+
+      k8s = {
+          enable = true;
+          node.enable = true;
+          master.enable = false;
+          inherit etcdCluster;
+        };
+      };
+
+    # services.kubernetes.kubelet.extraSANs = mkSANs {
+    #     name = cfg.name;
+    #     address = cfg.address;
+    # };
+        systemd.services.rc-local = {
+            description = "rc.local script";
+            wantedBy = [ "multi-user.target" ];
+            after = [ "network.target" ];
+            path = [ "/run/current-system/sw/" ];
+            serviceConfig = {
+              Type = "oneshot";
+            };
+            script = ''
+                # if [ -e /sys/block/md126 ]; then
+                #     echo "deadline" > /sys/block/md126/queue/scheduler
+                #     # echo "4096" > /sys/block/md126/queue/nr_requests
+                #     echo "4096" > /sys/block/md126/queue/read_ahead_kb
+                #     echo "always" > /sys/kernel/mm/transparent_hugepage/enabled
+                #     echo "always" > /sys/kernel/mm/transparent_hugepage/defrag
+                # fi
+                grep -q rdma /proc/fs/nfsd/portlist || echo "rdma 20049" > /proc/fs/nfsd/portlist
+            '';
+        };
+
+        boot.kernel.sysctl = {
+            "vm.dirty_background_ratio" = 5;
+            "vm.dirty_ratio" = 10;
+            "vm.vfs_cache_pressure" = 50;
+            "vm.min_free_kbytes" = 262144;
+        };
+
+        networking = {
+          hostName = name;
+          interfaces.enp33s0f3np3 = {
+            useDHCP = false;
+            ipv4.addresses = [ {
+              address = address;
+              prefixLength = 24;
+            } ];
+            ipv4.routes = [ {
+                address = "10.255.242.2";
+                prefixLength = 32;
+                via = "10.255.241.100";
+            } ];
+          };
+          interfaces."ibp65s0" = {
+            useDHCP = false;
+          };
+          interfaces."ibp65s0.7666" = {
+            useDHCP = false;
+            ipv4.addresses = [ {
+              address = "10.255.243.90";
+              prefixLength = 24;
+            } ];
+          };
+          firewall = {
+            allowedTCPPorts = [];
+            allowedUDPPorts = [];
+            extraCommands = ''
+              iptables -I INPUT -s 10.255.241.0/24 -j ACCEPT
+              iptables -I INPUT -s 10.255.243.0/24 -j ACCEPT
+              iptables -t nat -A POSTROUTING -s 10.255.243.0/24 -j MASQUERADE
+            '';
+          };
+        };
+
+        fileSystems = {
+          "/exports/work" = {
+            device = "/work";
+            options = [ "bind" ];
+          };
+          "/exports/data" = {
+            device = "/data";
+            options = [ "bind" ];
+          };
+          "/exports/opt" = {
+            device = "/opt";
+            options = [ "bind" ];
+          };
+          # "/vol/local-storage/vol1" = {
+          #   device = "/vol/vol1";
+          #   options = [ "bind" ];
+          # };
+        };
+
+        programs.singularity.enable = true;
+
+    imports = [
+        ../cluster.nix
+        ./hardware-configuration.nix
+    ];
+    };
+
+}