How to Seamlessly Connect Two Proxmox Servers for High-Availability Clusters

Published

Proxmox VE’s architecture thrives on scalability, but its true power emerges when you connect two Proxmox servers into a cohesive unit. Unlike standalone deployments, a linked infrastructure transforms isolated hosts into a resilient, load-balanced ecosystem—where VMs migrate transparently, storage replicates in real-time, and downtime becomes a relic of the past. The decision to bridge two Proxmox nodes isn’t just about redundancy; it’s a strategic shift toward operational efficiency, where manual intervention fades into the background and automation takes the wheel.

Yet the journey from concept to execution is fraught with pitfalls. Misconfigured networks can sever communication before it begins. Storage synchronization may lag under heavy I/O, leaving VMs vulnerable. And without proper authentication, the cluster risks becoming a security liability. These challenges demand precision—not just in setup, but in ongoing maintenance. The difference between a functional cluster and a fragile one often hinges on whether administrators anticipate edge cases before they materialize.

This guide cuts through the ambiguity. Whether you’re consolidating workloads, preparing for disaster recovery, or simply optimizing resource utilization, the methods outlined here ensure a secure, high-performance connection between Proxmox servers. We’ll dissect the technical underpinnings, compare clustering strategies, and address the nuances that separate theory from real-world reliability.

connect two proxmox servers

The Complete Overview of Connecting Two Proxmox Servers

At its core, integrating two Proxmox VE nodes revolves around three pillars: network unification, shared storage, and cluster management. The first step—establishing a robust network link—is non-negotiable. Without it, synchronization fails, and the cluster collapses into isolated silos. Proxmox leverages corosync and pve-cluster for heartbeats and quorum, but these tools require meticulous IP planning. Static IPs are mandatory; DHCP introduces instability. Meanwhile, the storage layer demands either a distributed filesystem (like Ceph) or a shared block device (iSCSI/LVM), where writes must propagate instantaneously to prevent data divergence.

The final piece is cluster membership, where each node registers with the others via a shared /etc/pve/user.cfg and /etc/pve/corosync.conf. This isn’t just about joining servers—it’s about defining roles: which node handles migrations, which acts as the primary storage target, and how failover triggers when a node goes offline. The process may seem straightforward, but the devil lies in the details: firewall rules that block cluster traffic, NTP misalignments causing time skew, or misconfigured voter settings that prevent quorum. Skipping these checks guarantees a cluster that works until it doesn’t.

Historical Background and Evolution

The concept of linking Proxmox servers traces back to the broader evolution of Linux high-availability (HA) clustering, which emerged in the early 2000s as enterprises sought to mirror mainframe reliability on commodity hardware. Proxmox VE, built atop Debian and KVM/QEMU, inherited this tradition but simplified the barrier to entry. Early implementations relied on heartbeat (now deprecated) and manual VM migration scripts, forcing administrators to script failover logic themselves. The introduction of corosync in 2010—paired with Proxmox’s native clustering support in 2014—automated much of this, but the underlying principles remained: a shared network, synchronized storage, and a consensus mechanism to elect a leader.

Table of Contents

Today, the landscape has shifted toward distributed storage (Ceph) and software-defined networking (Open vSwitch), which reduce single points of failure. Yet the fundamental challenge persists: ensuring that two Proxmox nodes operate as a single logical unit without sacrificing performance. The modern approach balances automation with granular control—allowing administrators to tune replication delays, adjust quorum thresholds, and even partition the cluster into sub-groups for multi-site deployments. This evolution reflects a broader trend in IT: moving from reactive redundancy to proactive, self-healing infrastructures.

Core Mechanisms: How It Works

The technical backbone of connecting Proxmox servers lies in three interlocking components. First, corosync handles node communication via multicast or UDP unicast, broadcasting heartbeats every second to detect failures. If a node misses three heartbeats, it’s declared dead, and the remaining nodes trigger failover. Second, the storage layer—whether via Ceph RBD, DRBD, or iSCSI—ensures data consistency through synchronous or asynchronous replication, with Proxmox’s pvesm managing the underlying block devices. Finally, the cluster service (pve-cluster) orchestrates VM migrations, resource allocation, and role assignment, using a distributed lock manager to prevent conflicts.

Under the hood, these components interact through a series of well-defined steps. When a new node joins, it authenticates via SSH keys (stored in /etc/pve/priv), then synchronizes its configuration with the cluster’s crm (Cluster Resource Manager) database. If storage is shared, the node mounts the replicated volumes and registers them with pvesm. Meanwhile, the network stack is validated: VLANs are checked for consistency, and firewall rules (iptables/nftables) are adjusted to permit cluster traffic on ports 5404 (HA) and 5405 (migration). The entire process is idempotent—re-running the commands won’t disrupt an existing cluster, a critical feature for automation.

Key Benefits and Crucial Impact

Organizations that successfully link two Proxmox servers gain more than redundancy—they unlock a paradigm shift in operational resilience. The most immediate benefit is high availability: if one node fails, VMs restart on the surviving host with minimal downtime. For mission-critical workloads, this translates to SLAs that would be impossible with standalone servers. Beyond uptime, the cluster enables live migration, allowing administrators to redistribute workloads without interruption—a boon for performance tuning and maintenance windows. Storage replication further reduces backup windows, as changes are mirrored in real-time rather than batched nightly.

Yet the advantages extend to cost efficiency. By consolidating workloads across two nodes, organizations minimize hardware sprawl while maximizing utilization. The ability to scale out—adding more nodes later—future-proofs the infrastructure against growth. Even from a security standpoint, clustering tightens control: centralized authentication via the Proxmox web interface simplifies access management, and quorum settings prevent split-brain scenarios where conflicting leaders could corrupt data.

— "A well-configured Proxmox cluster isn’t just a backup plan; it’s the foundation for a self-sustaining infrastructure where human error and hardware failure are mitigated by design."

— Proxmox VE Documentation Team

Major Advantages

  • Zero-Downtime Migrations: VMs relocate between nodes without pausing, leveraging shared storage and kernel-based virtualization.
  • Automated Failover: Corosync’s quorum system ensures that if Node A fails, Node B takes over within seconds, with no manual intervention.
  • Storage Redundancy: Replicated volumes (via Ceph or DRBD) eliminate single points of failure, protecting against disk corruption or node loss.
  • Centralized Management: The Proxmox web UI consolidates monitoring, logging, and configuration across all nodes, reducing administrative overhead.
  • Scalability: Additional nodes can be added later without disrupting existing workloads, thanks to Proxmox’s linear scaling model.

connect two proxmox servers - Ilustrasi 2

Comparative Analysis

Aspect Proxmox Cluster Alternative Solutions
Networking Model Corosync + UDP multicast/unicast; Open vSwitch for virtual networking. Pacemaker/Corosync (standalone), VMware vSphere HA (proprietary), OpenStack Nova (distributed).
Storage Options Ceph RBD, DRBD, iSCSI, or LVM over network (shared storage). VMware vSAN (software-defined), GlusterFS (distributed), ZFS-on-Linux (local replication).
Failover Time Sub-10 seconds for VM restart (with shared storage); sub-30s for full migration. VMware: ~30-60s; OpenStack: Variable (depends on scheduler).
Licensing Cost Open-source (free for commercial use under AGPL). VMware: $5,000+/node; OpenStack: Complex, often requires support contracts.

The next frontier for connecting Proxmox servers lies in hybrid and multi-cloud clustering. Today’s implementations are largely confined to on-premises data centers, but emerging tools like Proxmox’s pve-ha-manager extensions are paving the way for cross-cloud failover—where a Proxmox cluster in one region can mirror workloads to a cloud provider’s VM instances. This would eliminate the need for dedicated DR sites, reducing costs while improving recovery times. Simultaneously, advances in RDMA (Remote Direct Memory Access) networking promise to slash migration latency, making real-time synchronization feasible even across geographically dispersed nodes.

On the storage front, Ceph’s adoption is accelerating, with features like librbd integration allowing direct VM disk access without intermediate filesystems. This could further simplify Proxmox server interconnection by reducing dependency on traditional shared block devices. Meanwhile, AI-driven predictive scaling—where the cluster automatically adjusts resources based on workload patterns—may soon become standard, eliminating the need for manual tuning. The overarching trend is clear: what was once a manual, error-prone process is evolving into a self-optimizing, cross-platform ecosystem.

connect two proxmox servers - Ilustrasi 3

Conclusion

Connecting two Proxmox servers isn’t merely about adding redundancy—it’s about redefining the boundaries of what’s possible in virtualized environments. The initial setup may seem daunting, but the long-term dividends—fewer outages, lower costs, and greater flexibility—make it a necessity for any organization serious about infrastructure resilience. The key to success lies in meticulous planning: testing network paths before deployment, validating storage replication under load, and simulating failure scenarios to ensure failover behaves as expected.

As the technology matures, the barriers to entry will continue to fall. What once required deep expertise in clustering and storage now benefits from Proxmox’s intuitive web interface and extensive documentation. The future belongs to those who treat clustering not as an afterthought, but as the cornerstone of their infrastructure strategy. For those willing to invest the time, the payoff is a system that doesn’t just survive failures—it anticipates them.

Comprehensive FAQs

Q: Can I connect two Proxmox servers with different hardware?

A: Yes, but with caveats. Proxmox clusters require compatible CPU architectures (e.g., all x86_64) and similar kernel versions. VMs with hardware virtualization (KVM) will migrate seamlessly, but paravirtualized (PV) guests may fail if the underlying CPU features differ. Always test migrations between nodes before full deployment.

Q: What’s the minimum network bandwidth required for live migration?

A: Proxmox recommends a dedicated 1Gbps link for migrations, though 10Gbps is ideal for large VMs. Migration speed depends on VM size and storage type (shared storage reduces bandwidth needs). Asynchronous replication (e.g., DRBD) consumes less bandwidth than synchronous (Ceph RBD), but introduces slight data divergence risks.

Q: How do I troubleshoot a Proxmox cluster that won’t form?

A: Start with journalctl -u pve-cluster to check for errors. Common issues include:

  • Firewall blocking ports 5404 (HA) or 5405 (migration).
  • NTP misalignment (>5s skew).
  • Corosync misconfigured in /etc/corosync/corosync.conf (e.g., incorrect nodelist).
  • Missing or mismatched SSH keys in /etc/pve/priv.
Use pvecm status to verify quorum and ping -c 3 to test network reachability.

Q: Is Ceph mandatory for connecting Proxmox servers?

A: No, but it’s the most scalable option. Alternatives include:

  • DRBD (synchronous block replication, lower overhead).
  • iSCSI/LVM (shared block storage, but single point of failure).
  • NFS (simpler, but higher latency for writes).
Ceph excels in distributed setups but requires more resources. For small clusters, DRBD or iSCSI may suffice.

Q: How do I ensure VMs don’t lose data during a failover?

A: Use synchronous storage replication (Ceph RBD or DRBD) and enable pve-ha for VMs. Proxmox’s HA manager will:

  • Pause the VM on the failing node.
  • Resume it on another node with the same storage.
  • Synchronize any pending writes (if using async replication).
For critical databases, consider application-level replication (e.g., PostgreSQL streaming) alongside Proxmox’s HA.

Q: Can I mix Proxmox VE and Proxmox Backup Server in a cluster?

A: No, but you can integrate them. The Backup Server can manage backups for clustered VMs, but it doesn’t participate in the cluster itself. Use pbs backup commands to schedule snapshots, then restore to any node in the cluster. For disaster recovery, replicate backup data to a third location (e.g., another Proxmox cluster or cloud storage).

Q: What’s the impact of adding a third Proxmox server to an existing cluster?

A: Adding a third node introduces quorum complexity. With two nodes, the cluster tolerates one failure; with three, it can survive one node down but risks split-brain if two nodes lose contact. Adjust quorum_votes in corosync.conf (e.g., set to 2 for three nodes) and ensure all nodes have identical storage configurations. Test failover scenarios before full production use.