Big bang

06-form-cluster.sh

@@ -0,0 +1,72 @@
+#!/usr/bin/env bash
+# Form the 2-node cluster: create on pve1, join pve2. Sets up /etc/hosts and
+# root SSH trust between the nodes first so pvecm can run non-interactively.
+set -euo pipefail
+source "$(dirname "$0")/lab.env"
+
+HOSTS="10.10.10.11 pve1.${DOMAIN} pve1
+10.10.10.12 pve2.${DOMAIN} pve2
+10.10.10.13 pvestore.${DOMAIN} pvestore"
+
+echo "==> Ensuring /etc/hosts entries on both nodes"
+for ip in "$PVE1_IP" "$PVE2_IP"; do
+  lab_ssh "$ip" "grep -q 'pve1.${DOMAIN}' /etc/hosts || printf '%s\n' \"$HOSTS\" >> /etc/hosts"
+done
+
+echo "==> Establishing root SSH trust pve2 -> pve1 (needed by pvecm add)"
+# Ensure pve2 has a root key, then trust it on pve1; pre-seed known_hosts.
+lab_ssh "$PVE2_IP" "test -f /root/.ssh/id_rsa || ssh-keygen -t rsa -N '' -f /root/.ssh/id_rsa -q"
+PVE2_ROOT_PUB=$(lab_ssh "$PVE2_IP" "cat /root/.ssh/id_rsa.pub")
+lab_ssh "$PVE1_IP" "mkdir -p /root/.ssh && chmod 700 /root/.ssh; \
+  grep -qF '$PVE2_ROOT_PUB' /root/.ssh/authorized_keys 2>/dev/null || echo '$PVE2_ROOT_PUB' >> /root/.ssh/authorized_keys"
+# Pre-seed host keys so pvecm's ssh doesn't choke on prompts.
+lab_ssh "$PVE2_IP" "ssh-keyscan -H ${PVE1_IP} pve1 >> /root/.ssh/known_hosts 2>/dev/null"
+lab_ssh "$PVE1_IP" "ssh-keyscan -H ${PVE2_IP} pve2 >> /root/.ssh/known_hosts 2>/dev/null"
+
+echo "==> Creating cluster '${CLUSTER_NAME}' on pve1"
+if lab_ssh "$PVE1_IP" "pvecm status >/dev/null 2>&1"; then
+  echo "    pve1 already in a cluster; skipping create"
+else
+  lab_ssh "$PVE1_IP" "pvecm create ${CLUSTER_NAME}"
+fi
+
+# corosync reports quorate almost instantly for a single node, but pmxcfs
+# (/etc/pve) takes longer to become WRITABLE after 'create'. 'pvecm addnode' on
+# pve1 needs a writable /etc/pve or it dies "cluster not ready - no quorum?".
+# So wait on an actual write to /etc/pve, not on the corosync quorum flag.
+echo "==> Waiting for pve1 /etc/pve (pmxcfs) to become writable"
+for i in $(seq 1 36); do
+  lab_ssh "$PVE1_IP" "echo ok >/etc/pve/.joincheck 2>/dev/null && rm -f /etc/pve/.joincheck" && break
+  sleep 5
+done
+
+# 'pvecm add' can print an error yet still exit 0, so success must be judged by
+# the actual node count on pve1, and the join retried until it takes.
+pve_nodes() { lab_ssh "$PVE1_IP" "pvecm nodes 2>/dev/null" | grep -cE 'pve[12]' || true; }
+
+echo "==> Joining pve2 to the cluster"
+if [[ "$(pve_nodes)" -ge 2 ]]; then
+  echo "    pve2 already joined"
+else
+  for attempt in 1 2 3 4 5; do
+    lab_ssh "$PVE2_IP" "pvecm add ${PVE1_IP} --use_ssh" || true
+    sleep 5
+    cnt="$(pve_nodes)"
+    if [[ "$cnt" -ge 2 ]]; then echo "    joined: cluster now has $cnt node(s)"; break; fi
+    echo "    join attempt ${attempt}: still $cnt node(s); waiting before retry..."
+    sleep 10
+  done
+fi
+cnt="$(pve_nodes)"
+[[ "$cnt" -ge 2 ]] || { echo "ERROR: pve2 did not join (nodes=$cnt)"; exit 1; }
+
+echo "==> Waiting for quorum"
+for i in $(seq 1 30); do
+  if lab_ssh "$PVE1_IP" "pvecm status 2>/dev/null | grep -q 'Quorate: *Yes'"; then break; fi
+  sleep 5
+done
+
+echo "==> Cluster status"
+lab_ssh "$PVE1_IP" "pvecm status; echo; pvecm nodes"
+echo
+echo "Cluster formed. Next: ./07-add-shared-storage.sh"