Finally migrated off my 2016 OpenVZ box

Finally. After 4 years on a 2016 OpenVZ box at Hostinger, I am free. The node had 400+ day uptime which sounds impressive until you realize it kernel panicked weekly and the "uptime" was just because they never patched the host kernel.

My new home is a KVM at Contabo. YMMV, but in my experience the difference is night and day. IMO, OpenVZ in 2026 is a trap for anyone who needs actual isolation.

To everyone still on that Hostinger node: I wish you luck. Take it with a grain of salt when they say "migration planned."

If you don't mind me asking, what was the network setup? I've seen similar OpenVZ hosts using veth pairs with no proper L2 isolation, which creates interesting ARP issues when they finally try to migrate containers to a new host.

RPKI and IRR are irrelevant here obviously, but the BGP communities they advertised to upstreams—if any—would tell us whether they had any actual network engineering or just a single upstream with a /22 and a prayer.

Not that Hostinger ever published their IRR entries properly. I checked once.

To be fair, in my experience Windows Server containers on similar vintage hardware actually handle memory pressure better than OpenVZ ever did. Hyper-V isolation is genuinely decent even on older boxes.

I'm not saying Hostinger was good—clearly they weren't—but the container technology matters less than the operator competence. To be fair, most of these budget hosts would struggle regardless of hypervisor.

Glad you escaped, @quinnbizz.

Impossible combination. To be honest, you mean the guest uptime was fake, or the host rebooted without your container migrating?

As said, typical Hostinger behavior:

• Host kernel: 4.9.x or older
• No live migration capability
• Containers restarted on same broken node
• "Uptime" counted from container start, not actual availability

I left them in 2022. Same node, probably. As said, they do not invest.

Congratulations on your migration, @quinnbizz.

I moved my own stuff to Hetzner last year and this is what I found for users considering similar moves:

• Full KVM virtualization with dedicated kernel
• NVMe storage — no guaranteed IOPS but it's fast enough in practice
• Network: Dual-stack IPv4/IPv6, though I haven't messed with BGP
• No uptime SLA that I've seen, and maintenance windows are short notice sometimes

YMMV, but the price to performance is why I stayed.

(congrats (truly)) (though I am (still technically) on a 2015 box (at a different provider (don't ask))) (the shame) (my migration is "planned" (in the same way the heat death of the universe is "planned")) (so I am (in a sense) jealous) (also (nested parentheses (like containers (but worse))))

Thank you for choosing Contabo, Quinn! We're glad to have you with us.

If you (or anyone else migrating from similar environments) need any assistance optimizing your new setup, feel free to reach out via DM. I can check if there are any migration-specific configurations that might help.

— Jane @ Contabo