When Proxmox Server Solutions released the first alpha of the Proxmox Datacenter Manager — PDM for short — in the summer of 2024, community reactions were mixed. Finally a central tool for multiple clusters, some said. Others called it too late, too thin, and merely a reaction to the Broadcom shock that was pushing vCenter customers to look for alternatives. One year later — PDM has matured through several beta cycles and is now in production use at our customers — it is time for a sober interim review. What can the tool really do? Where is the gap to vCenter still wide? And for which customers does deployment already make sense today?
This article is deliberately not a feature fireworks show. We show what PDM actually solves in our daily work with SMB clusters in Bavaria, which expectations are realistic, and where a script, a second tool or a direct login to individual clusters is still required.
Why the Proxmox Datacenter Manager exists in the first place
A Proxmox cluster is designed to scale up to a few dozen nodes, sharing a configuration database — pmxcfs — and Corosync quorum. As long as you stay inside one data center, with one team, one permission structure and one network, the web GUI of any node acts as the control center. That has worked reliably for years.
However, as soon as multiple separate clusters enter the picture — two sites, a production and a DR cluster, a customer cluster and your own management cluster — there is no shared view. Every cluster has its own GUI, its own users, its own API tokens. Anyone who wants to move a VM from A to B works with qm remote-migrate on the command line, exactly as described in our article on live migration between clusters. Practical, but not something for colleagues without shell experience.
This is precisely the gap PDM is meant to close. An additional node — or a VM — with its own web interface talks to the individual clusters remotely over the familiar API, and bundles their states, resources and actions into a shared view.
The PDM update — practical status of what is stable today
We run PDM in several customer environments in production, always as a standalone Debian VM on one of the management clusters, with a dedicated HTTPS certificate and its own backup. Resource requirements are modest: two vCPUs, 4 GB of RAM and 32 GB of disk are enough to manage half a dozen clusters with several hundred VMs in total.
Multi-cluster overview
The first thing PDM gets right is the overview. After you have added several clusters through fingerprint and API token, a dashboard shows the number of nodes, cluster health — green, yellow, red — aggregated CPU and RAM usage and open alerts. For an administrator who used to keep five browser tabs open, that is a noticeable time saver.
The per-cluster detail view is deliberately kept lean. The intent is not to replace the native cluster GUI but to give an overview and offer a jump into the respective view. Anyone who wants to configure a VM is handed off to the cluster GUI — with a passed-through login as long as SSO is set up.
Central cross-cluster migration
The real heavyweight feature is central migration through the GUI. What used to be a command-line call with carefully constructed storage and bridge mappings is now a wizard: select the source VM, pick the target cluster, confirm the storage and network mapping, start the migration. Under the hood it is the same mechanism as qm remote-migrate, just with a proper interface in front of it.
In practice this works well as long as the clusters are technically compatible: the same or a newer PVE version on the target side, compatible storage types, sufficient bandwidth. We use it for hardware refresh, for reverse migration from staging to production, and for data center relocations. Compatibility details and limitations are collected in our ongoing Proxmox architecture series.
User and team management with SSO
PDM ships with its own user and team management that can be bound to existing identity providers via OIDC or LDAP. This lets you centrally define roles — read, write, migrate-allowed — that then act against the connected clusters as API tokens. While it is not yet as fine-grained as true delegation inside the clusters, it does cover the typical use cases in daily systems house work: first-level support is allowed to view and reboot, second-level to configure, architects to migrate.
Anyone already running a self-hosted SSO such as Authentik can connect PDM directly. For Active Directory environments, LDAP with the usual bind-user conventions works reliably.
Central API token and certificate management
A side effect of introducing PDM which we initially underestimated is proper token hygiene. Instead of creating API tokens for automation — Terraform, Ansible, monitoring — separately in every cluster, we can now trigger rotations in one place and keep an overview of which token holds which rights. That significantly reduces unpleasant surprises during audits.
What PDM still cannot do today — the gap to vCenter
As good as the basics feel, PDM is not a vCenter replacement with feature parity. Anyone coming from a VMware landscape should understand the gaps before marketing PDM as a one-to-one migration target. We collect the points here that keep coming up in customer projects.
No real Distributed Resource Scheduler
vCenter has DRS, a dynamic scheduler that automatically balances VMs between hosts. Proxmox added a Dynamic Load Balancer to the HA stack in PVE 9.2 — details in our PVE 9.2 release post — but it operates within a single cluster. PDM can see resources across clusters, but cannot move them automatically. Cross-cluster balancing remains a manual job or a script.
No deep per-resource RBAC
vCenter allows permissions all the way down to individual VMs, folders, resource pools and network segments. PDM currently operates mainly at cluster level. Anyone who needs fine-grained delegation per customer or per tenant still has to maintain the cluster-native ACLs or work with tags and conventions.
No central backup control
The Proxmox Backup Server remains its own target with its own GUI. PDM shows backup states and alerts in the overview but cannot start a backup job, trigger a restore or change retention. For a complete backup view, a direct login to the PBS — or monitoring — is still required.
No content-library equivalent
Templates, ISOs, snippets and Cloud-Init configurations still live per cluster. A central template catalog like vCenter’s Content Libraries is missing. Anyone unwilling to duplicate effort across clusters currently works with a shared ISO storage over NFS or S3 — a role that a TrueNAS system fills nicely.
No networking overlay management
SDN zones, VNets and firewall rules are defined per cluster. PDM can display them, but cannot orchestrate them across clusters. In practice that means anyone who wants identical network segments in both clusters either maintains them twice or deploys them through an IaC tool such as Terraform.
When PDM already pays off today
Despite the gaps, we see clear scenarios where PDM already delivers real value.
Two-site setups with a DR cluster
Customers with a production and a DR cluster benefit most. The central view across both clusters and the easy migration without shell gymnastics noticeably reduce the error rate during moves and test failovers. We usually complement the setup with Proxmox replication between two sites and a PBS in a third location.
Managed service providers with multiple customer clusters
When we as a systems house look after several separate customer clusters, PDM measurably shortens support reaction times. Instead of juggling passwords and URLs, the first-level colleague sees all clusters in one view and knows immediately where the problem is. For strict tenant separation — which of course we still enforce — a direct login per customer remains the clean solution.
Migration away from VMware
For customers moving away from VMware in the wake of the Broadcom reshape, PDM is an important argument in the middle of the pitch: the management tool does not have to be completely missing, it is just leaner. Anyone who honestly maps the vCenter functions and keeps only what they truly use daily can get very far with PDM plus the native cluster GUIs.
Operating, backing up and protecting PDM itself
A detail that is easy to overlook: PDM is itself a system that must be protected. If the PDM node fails, no cluster fails — the individual clusters continue to run independently — but the central view and GUI-driven migration are gone.
We therefore treat the PDM node like any other management VM: daily backup to PBS, configuration state in Git, documented restore procedure. For the underlying hardware we deliberately use separate resources so that a failure of the production cluster does not simultaneously take out the management tool.
For the storage backbone of the management environment a small, energy-efficient setup is sufficient — typically a TrueNAS Mini or an entry-level variant from our TrueNAS configurator. Large M-series systems are not required for the PDM role alone.
Effort for the rollout and what we realistically plan for
The pure installation of PDM is done within half a day: a Debian VM, add the package repository, install PDM, wire in the HTTPS certificate, register the first clusters. The real effort comes afterwards:
- SSO integration with a clean role matrix
- API token concept across all clusters
- Documentation for the colleagues so the new GUI is used and not the old bookmarks
- Monitoring for the PDM node itself
We deliberately do not quote fixed EUR prices here: the combination of hardware — often PDM simply runs on existing infrastructure — Proxmox subscription licenses and services is too individual. For a realistic estimate we are happy to provide an individual quote based on your existing clusters and goals.
Conclusion — PDM 2026 is grown up but is not vCenter
One year after its first alpha, the Proxmox Datacenter Manager is grown up enough for production use. Multi-cluster overview, central migration and SSO are stable and cover the most common systems-house use cases. Anyone coming from the VMware world must be aware of the gaps — DRS, deep RBAC, central backup, content library, cross-cluster SDN — but will be surprised how much daily work is already simplified by today’s feature set.
For the next beta cycles we expect progress mainly on fine-grained RBAC and on PBS integration. Until then the rule is: use PDM where it helps, keep the cluster GUI where it is better.
Frequently asked questions about the Proxmox Datacenter Manager
Is the Proxmox Datacenter Manager already released for production use?
Proxmox still lists it as beta, but in our experience it is stable enough for production use — provided that PDM itself is properly backed up and that a failure of the tool does not coincide with a failure of the monitored clusters. We recommend running it with a clear backup plan and a documented restore procedure.
Can PDM completely replace vCenter?
No, not across all functions. PDM covers multi-cluster overview, central migration and user management. DRS-equivalent scheduling across clusters, deep RBAC down to the VM level and a content library are missing. Anyone who strictly requires these functions has to close the gaps with other tools or conventions.
What hardware does the PDM node need?
Not much. Two vCPUs, 4 GB of RAM and 32 GB of disk are sufficient to manage several clusters with a few hundred VMs. In practice PDM runs at our customers as a lean Debian VM on a management cluster.
Does PDM support single sign-on with Active Directory or OIDC?
Yes. PDM binds to Active Directory via LDAP and to modern identity providers via OIDC. We often combine PDM with a self-hosted Authentik server, which works cleanly and is easy to maintain centrally.
Does PDM solve the topic of cross-cluster backup?
No. The Proxmox Backup Server remains the actual backup system with its own GUI. PDM displays alerts and states but does not orchestrate backup jobs or restores. Central backup management is currently not announced for PDM.
For which customer size does PDM pay off today?
PDM starts to make sense from two separate clusters onwards — typically production and DR — or at managed service providers with multiple customer environments. For a single-cluster customer, PDM does not yet deliver real added value.
More on these topics:
More articles
TrueNAS Proxmox Plugin exits pre-announcement: early-adopter rollout with signed APT repo
iXsystems has moved the TrueNAS storage plugin for Proxmox VE from pre-announcement to a concrete early-adopter rollout — signed APT packages for Bookworm and Trixie, a cluster-aware installer, multi-tenancy against a single TrueNAS, and official early-adopter status. What is new, what it means in practice, and where the limits are.
Proxmox licence costs vs. VMware and Hyper-V: an honest 2026 comparison
Proxmox subscription per socket, VMware VVF per core, Hyper-V Datacenter plus CALs: an honest 2026 order-of-magnitude comparison without fabricated EUR figures, based on DATAZONE advisory work.
Proxmox Backup Server vs. Veeam Community: Which When?
Proxmox Backup Server or Veeam Community Edition? Comparison of deduplication, VM limit, multi-hypervisor support and recovery workflows for SMB backup.