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.
What uptime should we expect out of our LEB?
Want to ask for opinion, actually what should we expect for the LEB we paid for?
Should we really expect for 99.99% uptime? Well, uptime is the most critical issue here.
As I heard BlueVM said for a few times, if you expect high uptime, go pay for dedicated. Although they say so, they do have a good SLA, and they do credit back to your account accordingly if there is a downtime.
BUT back to the question, would you really use LEB to host "mission critical" stuff?
Comments
99.9% excluding planned maintenance and outages with good reasons (hardware failure, node kernel updates, etc.)
no. i got kinda burned by corgitech just now when they needed to emergency reboot their shiz as i was in the middle of editing some stuff. but keep in mind most of these hosts (corgitech) are good about it.... hardware / network isn't magic. stuff goes down because .
Oh another question, for those "reasonable outages", especially hardware failure, anyone actually claim the SLA for credit? Sometimes I feel bad to claim it since im not paying much for it, and a few hours downtime most of the time render the whole month VPS usage for free... should I? Do they earn enough to pay back my money? Mine is hard earn money too :P
I don't, if the downtime is not so long.
Sure! Often the only difference between a LEB and a higher-priced VPS could be... only the cost. But of course a dedicated may provide you with better uptime (less factors to consider, not shared, etc). Most providers off a guarantee and/or SLA Credit Program which may be enough to re-pay you for your loss.
If you require 100% uptime, maybe a LEB isn't the best option. Look at dedicated servers: you see many reviews of various providers, their uptime isn't 100% either!
The path continues and brings us to a "high-quality", possibly more expensive, Dedicated provider which will guarantee 100% and has great reviews backing that up. <--If "mission critical". Etc, etc![:) :)](https://lowendtalk.com/resources/emoji/smile.png)
I'd suggest going for whatever meets you budget and looks promising and/or has great reviews.
Not sure this answers your question exactly but overall, I would expect from a LEB provider: What they tell me to expect (such as 99.9% uptime guarantee)
-Eric
If it was unreasonable downtime, I claim the credit.
Define:> @netomx said: not so long
In my experience...
2 LEBs in different locations = better uptime than 1 high end VPS (and usually a lot cheaper)
how do you loadbalance? DNS?
I use DNS failover at the moment (using 3 LEBs) but have been looking into other methods (i.e. 2 nginx or varnish proxies + 2 webservers).
Like, 3 hours?
lebHAProxy > leb1 > leb2 > so on
How much ram do i need for HAProxy? How low end can my haproxy box be?
And what happens if exactly the HAProxy box gets down?
100%. Why expect less than perfection? If my provider doesn't consider 100% uptime to be a goal, no matter their price, they are no longer my provider. You should never sell a product while saying to yourself "I expect my product to be consistently less than perfect." It's alright that it is, it's just not alright when you're totally satisfied with it and aren't constantly trying to think of ways to be better.
Why should you sacrifice uptime? 99.9% should be a industry standard. Unless some good reason for outage is available, provider should provide you guranteed uptime. The only thing that might/should differ between a LEB and HEB is price and spec. Not quality of the product.
My 2c.
LiquidHost :
DNS auto-failover onto second HAProxy - not hard.
As I say, loadbalancing the loadbalancers.
Obviously this is just taking things one step further and may be seen as being too advanced, but I find if you can afford it why not, its gonna be cheaper than most other options for high levels of redundancy.
@PAD
How do you sync mysql in real time? Remote mysql server?
http://rackerhacker.com/2008/01/03/mysql-replication-redundancy/
Converting a slave to a master if the master goes down is easily automated - when building your own redundant system you shouldn't have trouble coding basic PHP monitoring which can detect these things and do anything that needs to be done to solve a situation.
But it should be kept in mind that lowend providers often have very small staff, or even operate solo. That can affect response time and their ability to deal with multiple simulataneous issues.
This is the one I think many are overlooking. Everyone wants to pay less for anything, it makes us all happy, but then they expect a top of the line experience, and that just isn't realistic. You have to trade something to gain the lower price, and support is the most common trade off I see in the LEB space. Personally, I do not offer any SLA, I provide the service on a best effort, meaning I will put my best effort to insuring you get what you paid me for, and if something fails along the way, I will put my best effort into making you whole again. I do buy better hardware to insure my best effort is good enough to make most happy, but when I see thoughts about what SLA would you demand from your LEB provider, it just stinks of how entitled do you really think you are, should I be paying you to be my customer in your mind?
SLA? What SLA? This is LEB ffs
I don't measure uptime quality in % but in frequency of downtimes. Things can go wrong with every host and they will.. sooner or later. Hardware die, network issues, something... and if this happens twice or thrice per year, even if it takes more than hour to fix issue.. it's fine by me.
What I really dislike are regular downtimes, reboots, etc.. even if it takes mostly only minute or so to server come back online. Some hosts from some reason can't maintain even 2 - 3 weeks of uptime and this is what really annoys me.
In theory SLA is just a commitment to the customer that in event of downtime exceeding a certain time, customer will rebated at an pre-agreed rate. Yet to most customers anything less that 99.9% generally think that you are unreliable. To providers, we're just accounting possible failure and turn around time to get things fixed.
Personally, 99.9% uptime on any service that is reliant on a single point of failure is nothing more than a marketing gimmick. 99.9% means you can only have 43 minutes of downtime a month. If nothing goes wrong, 100% uptime isn't a problem. But when the worst happens (catastrophic failure), am I expected to believe that the provider can commission a new server and restore data within a span of 43 minutes? For VPS if you have a spare server around and your customers are willing to accept a clean VPS, sure that may work, but is that really the SLA customers are expecting? Even dell hardware replacement is generally within 4 hours.
Whether it's low end VPS, or high end VPS, as long as there's a single point of failure, customer expectations has to be realistic. LEB is a great idea if you're selling to the right people, aka server admins who (1) know how to optimize the server, (2) constantly take backups and (3) have a game plan when the server goes down to keep their servers running. Most LEB providers don't provide backups, prices are so low there's no way there's some redundant SAN backing the storage, plenty single points of failure. Most people who complain are expecting LEBs to have all the perks of a full VPS service but at dirt cheap prices. If nothing goes wrong, sure LEBs are a huge cost saving. But when something goes wrong they are the first to shoot at the provider because they were guaranteed 99.9% uptime and put all their eggs in 1 very low cost basket.
"Sorry for the downtime and complete loss of data, we'll rebate/refund you $5 for this month's downtime. I'm sorry for your loss of income of $10,000, web design cost of $5,000."
TL;DR:
(1) Realistic customer expectations, 99.9% uptime only covers 43mins to fix minor issues
(2) Backup backup backup, business continuity plan in event of downtime
I expect 100% uptime, minus maintenance and other unavoidable downtime. If a downtime event is completely avoidable but still happens (a single user abusing CPU and bringing the node down, the server running out of hard drive space, stuff like that which can be prevented), then I have a problem. Node needs a reboot for a kernel upgrade? Fine. Need to do maintenance on your rack's power strips? Fine. Just let me know ahead of time so that I can prep for it.
@NickM
Notices sent out by hosts for planned downtime is/should be a standard, so that's all great, but it should also be standard for consumers to work out their own plan when a LEB goes down. I find it pretty stupid that the customer expects the host to check all the boxes while they relax and sit back with ignorance - because even though the host normally will do everything in their power to avoid downtime, and it is standard to go through all of these procedures, it is NOT AN EXCUSE to slack on your own regime of avoiding loss, downtime, etc.
The standard should be applied both ways.
scheduled maintenance != downtime
If it brings customer services down then yes, it is. But it's "scheduled downtime".
hardware isn't magic.
That doesn't change that scheduled maintenance isn't downtime if it brings customer services down .