Case study · Manufacturing · VirtualizationStudi kasus · Manufaktur · Virtualisasi
Replacing oVirt, built right from the network up.Menggantikan oVirt, dibangun benar sejak jaringannya.
A manufacturing company in Indonesia needed a new platform to replace its existing oVirt virtualization. The answer: a 3-node hyperconverged cluster on Proxmox VE and Ceph, designed so storage traffic and cluster heartbeat never depend on a single path.Sebuah perusahaan manufaktur di Indonesia membutuhkan platform baru untuk menggantikan virtualisasi oVirt yang sudah ada. Jawabannya: cluster hyperconverged 3 node di atas Proxmox VE dan Ceph, dirancang agar trafik storage dan heartbeat cluster tidak bergantung pada satu jalur.
Summary for managementRingkasan untuk manajemen
- Why: the oVirt environment was running an outdated version, platform support had weakened, and the company needed hyperconverged storage that oVirt did not provide.
- Goal: a new, highly available platform for the company's servers, replacing the existing oVirt environment.
- Solution: three servers that each provide both compute and storage, running Proxmox VE and Ceph. VMs and lightweight Linux containers run side by side.
- Resilience: storage replication has its own redundant 25G network on two stacked switches, and the cluster keeps two independent heartbeat paths. If a node fails, protected VMs restart on another node.
- Capacity: 192 cores, 768 GB of memory and 45 SSDs of shared storage, with compression to stretch usable capacity.
- Alasan: lingkungan oVirt berjalan di versi yang sudah usang, dukungan platformnya melemah, dan perusahaan membutuhkan storage hyperconverged yang tidak disediakan oVirt.
- Tujuan: platform baru yang high availability untuk server perusahaan, menggantikan lingkungan oVirt yang ada.
- Solusi: tiga server yang masing-masing menyediakan compute sekaligus storage, menjalankan Proxmox VE dan Ceph. VM dan container Linux ringan berjalan berdampingan.
- Ketahanan: replikasi storage punya jaringan 25G redundan sendiri di dua switch yang di-stack, dan cluster memiliki dua jalur heartbeat independen. Jika satu node gagal, VM yang dilindungi berjalan ulang di node lain.
- Kapasitas: 192 core, memori 768 GB, dan 45 SSD sebagai shared storage, dengan kompresi untuk memperbesar kapasitas efektif.
Why replace oVirtMengapa oVirt diganti
Support around the oVirt platform had weakened, making long-term updates and fixes uncertain for production workloads.Dukungan untuk platform oVirt melemah, sehingga pembaruan dan perbaikan jangka panjang untuk beban kerja produksi menjadi tidak pasti.
The environment ran an outdated release, carrying the security and compatibility risks that come with it.Lingkungan berjalan di rilis yang sudah usang, dengan risiko keamanan dan kompatibilitas yang menyertainya.
The company wanted compute and replicated storage in the same servers, without a separate storage array.Perusahaan menginginkan compute dan storage tereplikasi di server yang sama, tanpa storage array terpisah.
Three networks, each with a jobTiga jaringan, masing-masing dengan tugasnya
In a hyperconverged cluster, storage replication is the heaviest and most latency-sensitive traffic. Giving it its own network keeps VM traffic and cluster heartbeat stable, even while Ceph is rebalancing.Di cluster hyperconverged, replikasi storage adalah trafik paling berat dan paling sensitif terhadap latensi. Memberinya jaringan sendiri menjaga trafik VM dan heartbeat cluster tetap stabil, bahkan saat Ceph sedang rebalancing.
Simplified diagram. Blue: management and VM traffic on VLANs to the existing core switch. Green: each node's bonded 2 × 25G Ceph links, one to each stacked switch. A separate 1G network (not shown) carries out-of-band server management.Diagram disederhanakan. Biru: trafik management dan VM dalam VLAN ke switch inti existing. Hijau: bond Ceph 2 × 25G setiap node, satu ke masing-masing switch yang di-stack. Jaringan 1G terpisah (tidak digambar) membawa out-of-band management server.
The solution, layer by layerSolusi, lapis demi lapis
ComputeCompute
- 3 nodes, each with 2 processors of 32 cores and 256 GB of memory.
- Proxmox VE installed on mirrored boot drives (ZFS mirror), separate from the data disks.
- Full VMs (Windows, Linux, Unix) on KVM and lightweight Linux containers on LXC, managed from one interface.
- 3 node, masing-masing 2 prosesor 32 core dan memori 256 GB.
- Proxmox VE dipasang di boot drive yang di-mirror (ZFS mirror), terpisah dari disk data.
- VM penuh (Windows, Linux, Unix) di atas KVM dan container Linux ringan di atas LXC, dikelola dari satu antarmuka.
Storage: all-flash CephStorage: Ceph all-flash
- 45 SSDs (15 per node) as Ceph OSDs, with monitors and managers on all three nodes.
- One production pool shared by all nodes, so VMs can move freely between them.
- LZ4 compression on the pool to stretch usable capacity.
- 45 SSD (15 per node) sebagai OSD Ceph, dengan monitor dan manager di ketiga node.
- Satu pool production yang dipakai bersama oleh semua node, sehingga VM bisa berpindah bebas antar node.
- Kompresi LZ4 di pool untuk memperbesar kapasitas efektif.
NetworkingJaringan
- Dedicated Ceph network: two 25G links per node in a failover bond, one to each of two switches stacked into one logical switch.
- Management and VM traffic separated by VLAN on 10G uplinks to the existing core switch.
- Out-of-band server management on its own VLAN.
- Jaringan Ceph khusus: dua link 25G per node dalam bond failover, satu ke masing-masing dari dua switch yang di-stack menjadi satu switch logis.
- Trafik management dan VM dipisahkan dengan VLAN di uplink 10G ke switch inti existing.
- Out-of-band management server di VLAN tersendiri.
Cluster resilienceKetahanan cluster
- Cluster heartbeat (corosync) on two links: the management network and the Ceph network. Losing one network does not break quorum.
- An HA group covering all three nodes: protected VMs restart automatically on a surviving node.
- Heartbeat cluster (corosync) di dua link: jaringan management dan jaringan Ceph. Kehilangan satu jaringan tidak memutus quorum.
- HA group yang mencakup ketiga node: VM yang dilindungi otomatis berjalan ulang di node yang masih hidup.
ResultsHasil
VMs and containers on hyperconverged compute and storage, replacing the oVirt environment.VM dan container di atas compute dan storage hyperconverged, menggantikan lingkungan oVirt.
Ceph replication runs on its own redundant 25G network.Replikasi Ceph berjalan di jaringan 25G redundan tersendiri.
Two corosync links and HA keep services running when a node or network fails.Dua link corosync dan HA menjaga layanan tetap berjalan saat node atau jaringan gagal.
Still running oVirt, or another platform you are not sure about?Masih memakai oVirt, atau platform yang masa depannya meragukan?
In a free 30-minute consultation we review your current platform, your growth plans and whether a Proxmox VE + Ceph cluster fits.Dalam konsultasi gratis 30 menit, kita tinjau platform saat ini, rencana pertumbuhan, dan apakah cluster Proxmox VE + Ceph cocok untuk Anda.
Customer identity, hardware brands, hostnames and identifying details are withheld for confidentiality.Identitas pelanggan, merek hardware, hostname, dan detail yang dapat mengidentifikasi tidak dicantumkan demi kerahasiaan.