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.
Comments
This may sound tough, but I do expect that my vps from any LEB provider should stay online and accessible as close as possible to 100%.
in my honest opinion, if you as a provider is happy with a 99.9% uptime, you are doing it wrong.
I do understand that many providers are a one man show (or only a few good men) but that i understand when I buy it. If I required 24/7 support I would look somewhere else.
I accept the TOS when I sign up that the vps is unmanaged, meaning that if I fuck up something in my os, its my responsibility to fix it. Or do a reinstall.
But its the providers responsibility to keep my vps accessible, aka online. Its also in the providers interest to keep their control panel online so I can reboot/reinstall whenever I like ( if this option is available)
as a customer I really don't care on how you do your job or make your profit selling "lowend boxes".
I'm just glad you are around for me.
The providers here that I have services with probably know that I dont send in tickets for every little thing. I always try to fix my own mess.
Bottom line is, I pay for lowend resources, not lowend quality. Which I have recieved from the providers I still have services with.
To be fair 99.99% is just an hour of downtime during a whole year which is absolutely fine.
@gsrdgrdghd then you already accepted that you will fail.
Instead of working towards 100% , I bet you are already coming up with excuses for that unplanned 1 hour downtime?
I understand that there are things that can happen thats out of the providers control. Nevertheless I'm sure many providers do every last thing to minimize those things.
If they are not, they will end up in trouble.
Murphy Murphy Murphy!
@MikHo
It is impossible to keep the uptime at 100%. The provider that offer 100% uptime cost around 10 times more than the most expensive lowendbox. Simply impossible to run on such a small profit margin and at the same time have enough money to have high-availability cluster. And in my experience, clients are satisfied with 99+ % uptime, so am I on every server I rent.
Hardware might always break so you cannot expect anything.
Hostbluff?
Personally I don't mind 99.9% or 99.8% as I usually have IP failovers or a backup ready to go, but what gets to me is how some providers handle downtime:
Situation:
Unknown downtime for an extended period of time, or fluctuating downtime (up 10 minutes, down for 10 minutes).
#1 - I expect to be informed that you're experiencing downtime. It doesn't have to be through email, but at least have a status page that one can refer to.
#2 - I shouldn't have to send a support ticket later that day to receive a vague answer of "there was a problem, it's fixed now".
#3 - NO NO NO. I do not want to follow you on Twitter to receive updates! (Post public updates if you use Twitter)
I've protested these points in the past with no avail. In general it's not asking a lot and could save you answering the same "downtime" ticket 100 times.
Even a hacked together status page would be beneficial:
I hear you but that is where the info is. I can show you were the water is, I can't help it if you refuse to drink it. I actually use twitter as little as possible, but if my website is inaccessible, I have to do something, and twitter is linked with the company facebook page so 2 birds.
I found "DDOS attack on our server" used frequently as one of the reason for downtime, do you think this reason is "reasonable"?
'reasonable' is in the minds of client, some may find it reasonable and some may not.
And those who find it reasonable would still want to know the steps taken to avoid the situation in future.
If the attacks tend to become regular then 'noone would be there' to find it reasonable.
i do expect 100% or 99.9% uptime, with proper intimation ahead of any maintenance.
As I say, loadbalancing the loadbalancers.
we'd love a How-To as separate topic
@liquidhost
Theres a huge difference between SLA and a providers own goal regarding uptime. I think you misunderstand my point.
And as discussed earlier, scheduled maintenance is an exception that will not count on the offered uptime.
I also think that the providers choice of hardware has a huge impact on the quality of service ( speaking with accessibility in mind, not diskspeed or vcpu cycles)
And if this hardware is monitored in a good way, problems usually shows in those logs before the "accident" happens.
And if you know there will be problems theres always a way to keep downtime to a minimum while fixing the problem.
Im in the business where we offer our customers a HA environment. We tell the customer that they can expect 99.99% uptime but our goal is 100%.
I always said LEBs are not for mission critical stuff.
That being said, since they are so cheap, you can always set up some redundancy and there are also some high quality providers that dont cave in under attacks, but s**t happens to everyone, some switch, some cable, some upstream failure, even hacks were not uncommon, so, back up often and put up some redundancy if you really need 100% uptime.
M
@joshuatly
define mission critical?
If I had a low resources application/site then why not? Its not that the Providers around here would give you any worse services then a high-end Provider.
Expensive providers fail too, downtime can be only prevented that much, expensive providers just have more budget to factor in redundancy. As a system(s) admin, if you know where the possible failures are, and plan for it, then it really doesn't matter if you're using a LEB or a Cloud service, or whether you're hosting your personal site or a mission critical site, maximum uptime can still be achieved.
Its similar to a shared hosting is what @Maounique is saying.
Lets say another person on the node was abusing the server in terms of I/O and bandwidth. Even if you're not getting the traffic, your speed and I/O can and will be affected by this. Or if someone decides to just crash the server (we're all thinking hypothetical people) and they just happened to get it to shut down, then your VPS is also down. If you want complete isolation then its probably best to go with a dedicated.
This is why I say its similar to shared hosting.
Mission Critical is also known as something that cannot possibly go down (if it goes down, everything goes down type deal).
Edit: and just like what @Kenshin said, everyone can/will fail at some point or another. Bigger companies just have more safety nets before they fall into the water.
@Pats
Maybe I will write one, but I would need to find a substitute for the software which I have created myself, since it is not something I am willing to release publicly. (Created a fully redundant H/A cluster which is fully automated and can be left alone with absolutely no interaction needed.)
As it was pointed out to you, thats bullshit. Downtime is time in which the service is inaccessible, no matter if its scheduled or not.
Instead of working towards 100% , I bet you are already coming up with excuses for that unplanned 1 hour downtime?
It is impossible to reach 100% uptime. Everyone knows that. So why should i try to achieve it? Also i don't run any critical servers and therefor i don't care if they are down an hour or two a year.
Same goes for a VPS with a High-End provider
But this can also happen if your going dedicated or with a High-End provider, I'm therefor saying that you can get as much uptime (and available resources) from a lowend provider as if your going dedicated or with a high-end provider.
If the provider do know what he/she is doing he/she can build the infrastructure to minimize the amount of single point of failures.
In the few times I've used support tickets my situation has been solved alot faster with lowend providers since the know their infrastructure on the back of their hand.
A High-end (larger) provider usually have staff where you need to send a few tickets just to make them understand the problem. (and the risk of not getting the same support personal)
That's true :P but usually high-end providers have more people or "eyes" to make sure the service doesn't mess up or generally have more quality equipment (this is all generalization and assumptions by the way) which usually (hopefully) reduces this.
Although the things you bring up afterwards do bring up valid arguments, but again that could depend upon the company and its internal structures. If you want a more personalized experience and one-on-one then obviously they could make it a single person at the same time, but in the market we're in its more focused upon getting the problem solved faster (within minutes/or a few hours instead of days) so that (getting the same person) really isn't likely.
Again I'm just saying, in my honest opinion there's just too many factors involved to generalize all large companies like that :P
My two cents.
Exactly, they have multiple routes, multiple switches, multiple nodes, HA or at least some automated back-up, etc.
With the safety net of a bigger provider with more resources, with own DC and personnel on call 24/7, it is likely that outages happen less frequently and last less than in the case of a one man show operation, which often does not have procedures in place for disaster recovery situations, sometimes at the cost of permanent data loss (yeah, back-ups, etc, but there is a limit in what they can achieve)
However, the degree of respect and support small businesses provide should be much higher, more communication, more custom plans, more attention to customer in general.
So it is a wide choice out there. Pick what suits you best.
M
@miTgiB I edited that right after posting (you must have seen the old version).
^ I don't mind if you post to Twitter, that's perfectly fine, but don't try to make me a "follower" just so I can receive updates. Those types of tweets should be made public.
I think providers should at least be able to match or do better than the uptime I get in my basement, as of yet none have been able to.
The almost 10 hours downtime, was scheduled maintenance downtime.
# uptime
2:33AM up 831 days, 1:02, 1 user, load averages: 0.07, 0.03, 0.01
# uname -a
FreeBSD mail 4.11-STABLE FreeBSD 4.11-STABLE #0: Mon Aug 6 13:25:24 SGT 2007 root@mail:/usr/obj/usr/src/sys/MAIL i386
Dell Dual Pentium 3 server at my primary datacentre. Still actively running email services to date but will be decommissioned this weekend.
Cisco 5505 also at my primary datacenter. 970 days uptime
well, we may see soon "disposable" machines. One run only :P
Install, leave it up, decommission. After 2 years and a third there might be a need for upgrade already.
M
I dont know...
something like if your site stop running for an hour or two you will loss couple hundreds bucks?
But you can argue that if you have that why use LEB right. haha.
actually im just asking.
I think I mentioned this before, the expected uptime should be what the provider claimed, be it 0.1% or 99.9%.
If the provider is unable to provide such numbers, I guess I'm OK as long as there isn't too much unscheduled downtime, probably < 1% or so.