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.
HA VPS?
I'm looking for a high-availability VPS. It will run a PHP script via CLI every minute, check some remote nodes, then make some API calls based on what the script determines.
Therefore, it does not need a lot of disk space, bandwidth, or memory, but I need it to be up and accessible > 99.99% of a month, or less-than 1 minute per week of downtime, including scheduled maintenance.
Any suggestions? I'm open to not-LEB-pricing, if anyone has suggestions for a product they've used before and can vouch for.
If I have to throw specs out there:
128mb of ram
2 gb of disk space
20gb of transfer
1 ipv4
Comments
Have you tried IPXcore? Ive never had any downtime with them.
Indeed, however some of the things that it will be checking are on IPXcore itself, so i'd rather have an "external" viewpoint.
what about all the "cloud" solutions
We use AWS (West Coast) for one node of our Nagios DNX installation
Aerosol water will not help
I forgot about AWS. Thanks for slapping me around, i'll look into AWS.
But i'm still open to suggestions otherwise.
DigitalOcean, 100% uptime since I've signed up (no I don't work for them and they did not pay me to post this)
Try http://www.zerigo.com/vps-servers I have been using then for over a year and have had 0 downtime including in their maintenance windows.
Hostigation in LA is very reliable:
root@la1:~# w 13:56:29 up 203 days, 5:20, 1 user, load average: 0.00, 0.00, 0.00
I tought there network was horrible because I had 3 random downtimes. Turns out your server gets a restart when you create a snapshot
Haven't had any downtime with them but have only been there for a few days. They are good tough.
Well.. Network outages do not infringe server downtimes..
They may be viable. While they have an SLA, and they like to plaster "cloud" on their website, they don't seem to effect the concept of HA.
Not HA. (unless i'm missing it on their website)
Not HA.
This was part of my requirement:
===========================
Have I discovered a new market?
I know.. But then it is just partially useful to post server uptimes.
And about the new market:
I guess some providers do this job quite well.
But at this market there is just barely any "HA", including a proper SLA.
Things go wrong. It is just as simple as that.
And as there is a lot of DDoSing around here and there, node abusing, DCMA takedowns and whatever might happen I'd just think about setting up two redundant VPS and keep them synced.
Just get two from different providers, close to HA as you can get. Such as linode and hostigation.
Well, DigitalOcean is partially cloudy, so try it
I've had a lot of bad luck with HA VPSs to be honest - 6 hours of unexplained and uncompensated downtime on Softlayer's Cloudlayer for example. I would just get something normal from a company you trust.
I'm using rackspace myself. Honestly haven't registered any downtime of their LBs and clouds the past half year, atleast according to my monitoring system.
James I see you are now with clustered. Moved from EuroVPS?
I'm not exactly sure what your definition of high availability is, but Digital Ocean has a 99.99% SLA (https://www.digitalocean.com/features).
That beats Amazon EC2's 99.95% (http://aws.amazon.com/ec2-sla/).
Im assuming N+1 redundancy on everything + auto failover.
An SLA does not guarantee this, all an SLA guarantees is you receive credits when there is downtime.
Have you considered running it from multiple locations in a pseudo-roundrobin set-up?
Server X runs the script at 00:01
Server Y runs the script at 00:02
Sever Z runs the script at 00:03
etc.
If the PHP script is portable, then by increasing the number of servers, you're simultaneously decreasing the likelihood that all three will be offline at the same time - Even if two are down the script will still run, albeit in a degraded state.
I wouldn't trust too much on the auto-failover of EC2 instances
You are much better off building your HA config at the app level, unless you have control of the hardware
genius
Yeah. EuroVPS was actually fine while I was there but they've had some SAN-related downtime today.
Hopefully Clustered will be the last move for a while, though effectively I'm paying £50/month to host 1 site which is disappointing (because of the resources DeskPRO + R1soft eat mainly...)
Clustered use local storage but everything's redundant all the way from the network to the power to the BIOS (and the performance is absolutely amazing).
Anyway back on topic I would ask @Damian to consider my experiences before jumping on the HA bandwagon. HA is great on paper but when it comes to uptime figures, I think you can do just as well with any quality provider.
Uncle has special HA setups for his corporate customers using vmware and rhev. I think the price wont be LEB, tho.
Maybe that's the case with smaller hosts/LEB providers, but for large infrastructure services SLAs aren't primarily about refunds -- they're about displaying how confident a provider is, or isn't, in their own ability to provide a given level of service. If a company (at least, a reliable one) advertises 99.9whatever% uptime through an SLA, the general expectation is that you'll get that uptime.
Large infrastructure SLAs are contract based; however at the end of the day come down to the same thing.
The only SLA to trust, is one based on historical data over a long period.
I have gotten downtime on my NY VPS the last few days. 5 minutes both times.
Edit: I just looked at my pingdom and Icinga reports and I have 8 outages. 21 minutes. 99.79% so far.
NL location seems better so far.
An SLA is legally a contract if there are financial penalties for missing an absolute level of service.
Everything which is made by humans will fail eventually, Murphy always finds ways to break it that you never imagined.
It is mostly down to luck, taking all the imaginable steps to reduce the risk from 99.5 to 99.95 will make your costs 10-20 times higher and nobody will wish to pay for that service, it is better and safer to make your own HA scheme with many providers.
Regarding SLAs, anyone can offer an SLA. Consider that most providers are never down for days; it's tuppence to give anyone a few cents of credit for an hour or two of downtime and a "sorry!" note than to implement any sort of system to actually ensure that things are up. My point of not being very interested in a company that offers SLA is that i'd rather have the server actually be up, rather than have it yo-yo up and down and get less than a dollar in credit.
SLAs might make sense when you get into the $1000+ range (because an hour of downtime at $1k/month+ hurts the provider more than a $10/month product), but below that, they're just blowing smoke up your skirt.
Thank you (and anyone else I missed) for your input; it appears that this would be the best idea anyway. Maybe some sort of voting system with an odd number of nodes.