All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Looking for KVM providers of high system uptime
Hi,
I am looking for recommendations of KVM providers that provide very high uptime. By uptime, my concern is mainly the operating system, not network uptime. I find it's difficult to achieve high os uptime: hardware failure, datacentre issues, bleeding edge software, fancy tweaks that break the host os, bluffing cpu bugs being discovered...
Because I have a lump of services that I need to start manually after os reboots, and there's no easy init scripts that automatically bring them up online (and I don't really bother writing either). So I am looking for some VPS that stays up for like a year between reboots. Network uptime is not a big concern, hence how many 9's of % uptime doesn't really matter as long as it's reasonable. Security is not a concern either, as I'm counting on providers not aggressively applying every patch out there that fixes some holes that are not remotely exploitable in reality.
By far my Virmach KVM has the highest uptime, almost a year as of now. But I'm looking for other options in the low-end range. Any recommendations are appreciated.
Comments
Take a look at hostdoc (gold Xeon) and ultravps.
I have a storage VPS with SpeedyKVM:
You should however bother with automating your stuff, regardless of how good the uptime is.
CrownCloud comes to mind
Also ionswitch
Many others on LET are pretty good (comparable with Virmach) - could easily name another dozen (and +1 for the ones already mentioned above) - maybe you might narrow down your options a bit by preferrence for location etc ?
Hetzner and OVH and probably others have "cloud" VM's with network file systems (e.g. CEPH), that can migrate automatically between host nodes in the event that a node crashes. I don't know if a reboot is visible to the user if that happens. Yes if you're doing anything serious you need to automate your restart
Just pick any decent provider, had more than a year uptime with Hetzner, Netcup, HostHatch, UltraVPS, more than 3 years with Iniz (ended up cancelling the server).
It's 2019, there is certainly a viable easy enough solution if you look into it.
Yeah, I know, for most services. But mine's some half-ass pieces of scripts that may result in data inconsistencies between reboots, which require manual intervention. The cost of fixing or rewriting the software is probably much greater than years of some low-end VPS.
Thanks, location doesn't really matter for me. Could you name another few that you have experience with?
Thanks for all the mentions. In fact, I asked around before posting, but even some 'Top providers' on LET are not that good at keeping up the uptime. They might have 99.99% uptime, but many of them had scheduled and unscheduled reboots within the past year. They are good at keeping the downtime minimum, but that's not what I'm looking for...
Then, LET providers are not suitable for your mission critical workload. Perhaps consider Softlayer, AWS etc.
Here's a couple of our hypervisors:
[tab.fitts@kvm-3-us-tx ~]$ uptime
11:18:42 up 412 days, 21:11, 2 users, load average: 4.94, 4.92, 4.81
[tab.fitts@ssd-kvm-4-us-tx ~]$ uptime
09:18:39 up 410 days, 10:23, 1 user, load average: 5.37, 5.40, 5.15
We use online kernel patching (KernelCare) to maintain security patches.
mmm ... thinking about what you're asking for, I don't want to be too glib making so many suggestions without studying a bit more carefully
as @SpryServers_Tab mentioned above, KernelCare (or similar live patching capability) will be a necessary ingredient unless you like hanging out on an unnecessarily vulnerable system (which I try to avoid, so really can't recommend any along those lines).
other important factors to consider might include A+B power, quality hardware (newish yet well-tested, not surplus), competence in system configuration (wherewithal in designing effective RAID etc redundancy being the tip of the iceberg), and strong system administration experience
And of course buying something advertised to be all of the above is still no guarantee of 100% uptime - if that's what you need, then it's ultimately going to be up to you to either design for it ("high availability" etc) or pay a bit more for that kind of SLA from someone who makes it their business to provide just that.
tl;dr - you pays your money and you takes your chances. @SpryServers_Tab seems to be bringing receipts for 400+ days uptime on their nodes, and though I don't have experience with their stuff I'm inclined to say it looks pretty good. And you might also take a closer look at some affordable yet experienced providers here such as: AlphaVPS.bg, First-Root, and Lunanode. They are active on LET (AlphaVPS -> @AlexBarakov / First-Root -> @FR_Michael / Lunanode -> @perennate
) and I mention them because in my experience they are transparent, no BS people to deal with who will be happy to tell you to look elsewhere if they're not right for your application.For example - reading https://status.lunanode.com/ will give you an eyeful of gory details - and yet my $3.50 KVM instance with them in Toronto seems to be pretty solid in terms of uptime.
Plenty others that "work great for me" but given your specific requirements maybe better to focus on a few and narrow it down.
He wants close to perfect uptime, not pretty good. What he asking for is HA and redundancy to max for the LET price. Sure some providers can give you stats of years of uptime, but tomorrow is another day and one bad KernelCare patch and that is gone and not what he wants.
He does not care about about network, which means the app in question here does not require it. Not sure what the app might be, nor do I care really, but for what he looking for it should not fall into the LET category. This sounds like specialized hosting and with that, comes money.
what OP can consider is Tier 4
https://www.massivegrid.com/high-availability-dedicated-cloud-servers/
In my experience with some openvz hosts is that if they have to reboot the hypervisor for whatever reason, the container is simply suspended before the hypervisor goes down and resumed when it comes back up - the end result being you don't need to restart your application stack.
In my opinion, do whatever it takes to automate that shit. It's worth it in the end, and you will gain some very valuable experience and knowledge that you will for sure find useful in the future.
Point is he does not want to automate anything. He wants the provider to have burden his problems. Then he needs to pay the provider to burden his problems and not bring this to LET.
Heh, well if that's the case I wish him all the best..
There's probably a flip side too, to be honest, I have one service I don't want to automate (requires me authenticating in a certain way by design) but it would certainly be possible if security was less of a concern.
Edit: we can both agree OP should hire a good sysadmin.
Oh I have several that require me to intervene, but I take that upon myself and do not ask anyone to do it for me. You understand as well, it can be done, but should not be done type of thing.
His app does not require the net, thus he could just have ssh access, I assume, and automate it. He would have no security reasons to have any port open but ssh, and not even that if he used a VNC from a provider. Yes I know supposed to be for rescue and what not, but could be done.
If you lock down a server to that point, then you have no open facing surfaces for attack and thus security becomes a second thought. The only network into the server would be through the provider and if some one bunks your provider you would screwed anyways.
I've had great uptime with VirMach.
Virmach was the best provider on CC network as long as I know. At least he was longest provider on that network.
You don't need to write init scripts, just make a shell script put your commands in it and add it as a cronjob.
https://www.cyberciti.biz/faq/linux-execute-cron-job-after-system-reboot/