New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
How a VPS Provider ensures Data Redundancy of the Nodes ?
Mahfuz_SS_EHL
Host Rep, Veteran
Hello,
Here at LET, I find many VPS Providers. Most of the providers use SolusVM, some use others. Also KVM Virtualization is mostly used.
May I know for redundancy of Data what sort of steps you take ??
At Drive Level, you utilize Raid-1 or Raid-10 ??
And then again, keep backup of the VMs in Daily/Weekly manner ??
What about the NVMe drives ??
I'm basically asking what sort of protocol you use for redundancy & disaster recovery.
Regards.
Comments
I personally do backups of web hosting customers, but I don't do that for GPU compute customers because the local storage should be temporary until an ML model is trained. Network-based storage with replication across disks (in clusters) is where long term data should be held.
One of my friends at a "big 3" cloud told me he once had a customer who had a VM on a host server that crashed and the data was lost because the customer didn't implement any backups. That VM sent daily reports to the CEO of a Fortune 500 company. So even "big 3" clouds likely don't have much redundancy, probably only replicated storage and no backups except for larger customers / more expensive services.
Our Antarctica IPv9 VPS utilizes highly durable CF cards as storage.
There's no RAID or redundancy or backups of any kind.
Always build redundancy at high levels, so that you don't rely on a single storage device or a single server.
Alternatively, YOLO live dangerously and believe in the end.
For us, we have centralised share storage via FC, the storage array with its own raid redundancy with hot spares at the ready.
We use proxmox cluster for all our virtualisation which means HA comes out of the box for both containers and VMs.
If one node goes down for what ever reason the VMs get booted up instantly on the other nodes (automatically distributed based on load/rules)
This does mean the VMs have a very brief outage while they boot (sub 60seconds)
The other advantage of the proxmox clustering that if we ever want decommission or take a node offline we can do live migration of our VM to other nodes without any VM downtime emptying the node from all customers before doing maintenance/rebooting/decommissioning the node
Yup. If you're not paying for backups specifically, you should expect they're not happening. RAID is nice until a controller goes haywire, and while that doesn't happen often it only takes one out of a million to be one of the most repeated stories around the internet.
So, those who don't utilize any sort of HA/Cluster or Network Shared Storage don't provide any level of redundancy right ?
Typically no, outside of RAID. And even those I wouldn't rely on. I've heard of more SAN failures than RAID10 failures. There is no situation where you shouldn't be doing backups. I've seen data loss on giant ceph clusters. I've been the one breaking the news.
I don't think that's true. You can have local redundancy i.e. raid1 raid5 raid6 or any of the 0 variants; and not have HA etc. Don't use raid5 though.
However, you can also have HA and make that your single point of failure. I've seen a lot of stories of Ceph getting involuntrated. Hostsolutions are a good example of what happens when a large HA cluster dies - which is exactly the same as when locally redundant storage dies, but more people are impacted.
Ultimately if your data has value to you, take your own backups. Providers can use/offer whatever levels of redundancy they like but if you don't back up your data and verify your backups there's absolutely no guarantee unless agreed in your contract, and even then that's down to legalese.
Back up your data.
He's also seen attack ships on fire off the shoulder of Orion and C-beams glitter in the dark near the Tannhäuser Gate. And just like your data, all these moments will be lost in time.
(btw I like this related song)