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.
More Reliable VPS Needed
I've been using my current vps provider for about a year now, and it has been decently reliable so far, but it has brief bursts of downtime each month, such as last night when it was up and down for 6 hours.
The site I am running on it has gotten more popular, and it has gotten to the point where it makes financial sense to pay for a more reliable server. What is recommended? Is Amazon EC2 more reliable?
Comments
Yes.
I guess my problem is that I see no easy way to compare provider reliability. Something like Linode also looks interesting, but it seems painful to pay a factor 10x more just to get a little more reliability.
I'd recommend you get a couple new VPSs from reputable providers and have some form of failover between them.
For example:
If you wanted, you could add a third layer into this for extra redundancy, but this alone should protect you from downtime in most cases.
I would suggest Rackspace over EC2 any day. Main issues i had with EC2 were with regard to minimal performance and lack of proper support , hence moved to Rackspace
We have a few Rackspace VPS at my place of employment for various widgets and services for use across a few sites, and have had no issues, network outages, or unexpected downtime in the past two years.
I'm leaning towards something like Linode right now ($20 a month). The price is a little bit difficult to digest since I am paying under $2 a month right now, but I'm making enough off the site that the additional cost makes sense for more reliability.
@ElliotJ I also have been thinking about doing this. My site is a php site that runs off a mysql database. The database is constantly being updated by a separate process, but the state of the database is not critical. Basically, I could have a master node that holds the main database (that is updated by my "separate process"), and then replicate the database out to my other nodes that I can switch to in-case of downtime.
What's the best way to replicate a mysql database between servers? Basically I want a master node, and then multiple slave nodes. The slave nodes should always match the master node, but none of the slave nodes need to have write permission. If the master node goes down, I can run off one of the slave nodes (without writing to it) until the master server comes back up.
NO
You should consider buying offloaded MySQL.
And it is not critical for my slave nodes to be 100% up to date. If they are a few hours behind, it's no big deal.
I think this is probably the cheapest option, but it's going to be the most difficult to implement. I should have some free time over the holidays to attempt to implement something like this. My only concern is that some web crawlers seem to cache dns ip's more than they are supposed to leading to crawlers accessing the wrong server (I noticed this when I switched servers a few months ago).
@Spencer Why? Specific story?
@gubbyte What would offloaded mysql gain me? That just gives me an additional service that could go down, and make my server stop working. It seems like that would make it worse.
That's true. Nevermind then; disregard my post.
I've recently been playing around with EC2 and the performance really is crap on Tiny and Small, however, you get a free RDS instance for MySQL, so you can have that do your processing and just use the EC2 instance as a webserver.
Have a look at Percona, there's plenty of documentation on their site as well to guide you through the process.
Why not just pay like $13 a month for a mKS-2G dedicated server from kimsufi?
Because he's looking for reliability which can be easily obtained on a VPS from the right providers. Also a single dedicated server is still a single point of failure, therefore when/if it goes down your site is still down. We're looking for reliability not power here.
Linode has decreased (in my opinion) in quality over the past few months and now isn't really as reliable. They also have a history (here) of horrible customer support and just have the extra price tag for being a "premium service".
Personally, I'd go with an alternative approach (not EC2 or Rackspace) similar to what @ElliotJ stated. Have like two or three VPSes with automatic failover (with Rage4) and having them rsync to the main node.
@curtisg That's actually pretty tempting, but I'm not sure how the performance compares to a vps that is hosted on a multicore machine (or if I need to be worried about that - my application really isn't cpu taxing).
Also, does anyone know how the network uptime / machine uptime is on the Kimsufi machines? That's my biggest concern.
My kimsufi machines (three) have not been down ever. I have received 0 issues. Network or server.
@HalfEatenPie Agreed, the distributed approach is probably my best bet, I just need to do some testing to see how reliable / easy to set up it is. I'm fine with spending several hours setting it up, as long as I don't have to spend a lot of time on maintenance. If it becomes a maintenance nightmare, paying an extra $10 a month would be worth it to me.
Thanks for the info on Linode. For $20 a month, I would expect them to have an easy to use control panel, and be highly reliable (almost 0 downtime each month), considering I would be paying close to 10x more than I am now.
@xur17: FreeVPS.us is on a $15/year VPS and the owner @dmmcintyre3 makes 10x that in a month off of the site alone. It doesn't go down very often (if ever, to my knowledge, of course I Don't go to that site regularly so who knows).
Honestly, what you're trying to compare is the cost -> Uptime, but what I'd like to let you know is Cost =/= Uptime. The provider and your own settings are what factors into the uptime.
Look on this forum, talk to the providers (they're more than willing to work with you for your business!) and make the best decision for yourself. I can recommend a ton of providers that I'm currently with, but in the end its how you want to execute/get to your plan/destination.
Throwing more money at a problem works only 10% of the time.
@HalfEatenPie Kimsufi might be a good option for me then. For an extra $10 / month I get 8x the amount of ram I have now, higher reliability, and more storage space.
I might be overstating my reliability needs - I'm fine with a few minutes of downtime a month, but I had more than an hour of downtime with my server yesterday, and lost all data on my server a few months back (leading to close to a day of downtime). I had backups for the server, so I was fine, but that led to close to a day of downtime, and quite a bit of time spent re-configuring my server.
I'll just come out and name my provider (I was hoping to avoid doing this). I've been with buyvm for close to a year now, and everything I read when I joined them seemed to indicate they were great and highly reliable. The more I use them though, the less it feels like that is true.
Basically I see two options here:
1) Switch to a more reliable provider, and possibly pay more for it. The downside to this is that I have no guarantee of the quality of the provider, I can only base this on other people's experiences. This is the simplest option assuming I choose a good provider to switch to.
2) I can create a multi-server setup, and use a setup to automatically fail-over to a different server when I have issues. This will take a good chunk of time to setup, and possibly will be painful to maintain.
Know this, Kimsufi brand itself isn't 100% reliable. You're getting outdated equipment (sometimes 3 - 4 years old equipment that have been run to the ground by their previous users/owners. They're literally put out there till the end of time and once they break they put in a new-used one. This means that your files/server could go down at any time due to hardware failure. While OVH might be great/awesome with replacing the failed hardware, if/when it happens you'll have a big headache to deal with (especially with the possibility of all your files just vanishing). Kimsufi comes with no guarantees that your files will remain intact during use. I have 0 mission-critical files on my Kimsufi dedicated servers, and I will keep it that way due to that reasoning.
Also, with OVH and Kimsufi, their support staff (hardware wise) is fantastic. They'll send a support staff within an hour or so to take a look at your server HARDWARE wise. But, if you suddenly nuke your server and it goes down, then its all on you to fix it (they won't touch it). They're seriously 100% THE Unmanaged service. The automated console will give you basic access (boot to rescue mode, etc.) but that'll be it. You have to fix everything yourself if/when it goes down. (This happened to me a while back, funny/stupid story, people on IRC know it, but its not fun to deal with). Also realize their network is just there. It works, its not quality, but it works.
BuyVM personally has been fantastic to me. I do know their New York location is still falling a bit short in comparison to their San Jose location, but San Jose has been perfect. I don't know specifically what location you're with them, but I would ask you to reconsider due to my personal experience with them.
I'd say talk to providers directly. Everyone will say they have great uptime, but its what else they bring to the table that matters. (This is the part I start naming off providers that I know/use) I know for a fact @KuJoe goes crazy with backups (I'm not too sure if the customer's VPSes also factor into it but its worth a shot asking him) and his customer support has a ridiculous response time. IPXCore has been great with uptime and working with me/other people on their needs. Hell everyone listed here have great uptime and respond to support tickets pretty fast. Seriously @Damian is one top notch guy. @miTgiB's the "one-man-superman-team" behind Hostigation (well, now two-man but seriously, Tim knows his things). @FRCorey and his Colorado location (KVM VPSes with him has been great) has a great network/system running. Also he definitely knows his things. @Jarland and CatalystHost, I swear every network/downtime e-mail I get from him (not that I'm saying I get a ton, just saying those that do come) he's probably one of the most down-to-earth guys and cuts right to what we want to hear: What went wrong, how he's working to fix it, and what he's going to do in the future to prevent that from happening. Obviously EDIS falls in there for anywhere else in the world (but they're only vServers and KVMs). Of course, GetKVM, my status/network monitoring script has had 0 downtime (alteast noticeable for me), with @GetKVM_Ash. But of course location does matter to where you're located/your focus group is.
There's obviously more providers out there that I absolutely approve of (sorry if I blanked ya!), but this hungover pie just can't think right now (also I'm procrastinating on some work).
With the tools you have available, I don't see how 1) and 2) can't be combined. There are many tools out there including Puppet server management scripts (there should be a LowEndTalk thread about this, I don't remember where)) and for the fail-over the only thing else you'll need is Rage4.com's DNS services.
Rage4.com you'll only need to update when you're adding/removing new servers (or whatever DNS issues you have), and the puppet/slave server management software can be used to update your VPSes. Individual SSH use can be done when a problem arises that's only focused on one VPS.
You can also set up monitoring scripts (Like Nagios or Munin) which will let you know if any VPS is down/having issues.
The biggest milestone in this project isn't really the maintenance, maintenance should be a piece of cake, it should be the setup of the project.
This is a rumour and it's totally wrong.
For reliability, I would suggest FitVPS and PogiWeb (among LEB providers).
But to go a step further, you need two VPS in cluster (with mysql replication), DNS with low TTL and scripts to monitor and update automatically DNS. You can do that for $20/mo.
Take a linode for yourself if you talk about reliability.
ammm - StyleXNetworks is reliable too. And cloud3k.
Its been a hit or miss for me personally.
DigitalOcean. 'nuff said
https://www.digitalocean.com/blog_posts/jsfiddle-net-moves-to-digitalocean
Google for MySQL master-slave replication.
Have you ever been a customer of them then?
That's with all dedicated server providers, besides HDD nothing can be permanently harmed really...
Sometimes hardware fails though (CPU mainly) but your data will be safe.
Sometimes hardware fails though (CPU mainly) but your data will be safe.
From personal experience (atleast), hard drive failure has been the bigger culprit for me
I read your thing wrong. Yeah.
Their control panel is the best in the business, IMO.
They are highly reliable, and their technical support is next to none. I've been a Linode customer for about 5 years, and they are just as good today as they ever were in the past.
If you want to do some kind of distributed setup using LEBs, by all means go for it. If you're comfortable paying the Linode price, don't hesitate. And paid annually, it's $18/month....
Linode is perfectly fine if you have the budget, they also have started to deploy newer hardware and uptime is generally 100%. Also as mentioned above the custom control panel will do everything you need and more
http://www.linode.com/linodes/