And it does beg the question if they can't manage their own infrastructure why would anyone be silly enough to let them manage theirs ?
I'm pretty sure I'd never trust Linode even as just a bare VPS provider, let alone giving them more access to my machines to provide this support, though, given their long and horrible track record.
I'm genuinely curious. I've been hosting with them for a couple years and have had zero problems and amazing support response rate otherwise.
If there's some kinda of a nightmare scenario waiting for me, I'd love a heads up.
Also their support is only good for minor incidents. When you have a data centre go down their support evaporates and you will be left with a multi-hour outage with no clue what is happening.
Is that still the case?
AWS networking continues to kind of suck; Google probably does at least that part better.
Other options might be Rackspace, Windows Azure (I think you can run Linux VMs, however I dont know how well that's working) or Google Compute Engine (rather new, so YMMV). Those are probably the biggest players in the cloud market.
Why? Too much "scheduled downtime" (few hours/month) vs. AWS's 99.999% uptime.
Why else? Well, it's a perverse incentive. If they have less reliable systems, the pain goes up, to where more people will sign up for "Linode Managed". We should all get good uptime, not have to pay extra.
Most importantly, their offering is not an SLA. I don't see anywhere in the "Linode Managed" where they are guaranteeing uptime %'s, or penalties for lower performance. And, how can they realistically handle even 50% of the 3AM panic problems, if some of those will be my website's inability to talk to 3rd party sites (which neither of us have control over).
If they really can fix 90% of the 3AM problems, then they are the root cause of most of those in the first place. There's not many good reasons why Linux will break at 3AM often if you've setup your stack correctly.
My Linodes are rock solid. Yeah Linode got hit by a cold fusion vulnerability and had the bitcoin issue, but for company as old as it is, that's still a good track record imho.
Meanwhile we're leaving EC2. Don't need it really and certainly don't need the headaches. The CPU on those things are pretty weak as well.
The grass isn't always greener, and with so many users, our personal experiences are simply "statistical anomalies / anecdotes" to the other guy, but are 100% our relevant personal experience to ourselves.
It kind of reminds me of my uncle who does surgery on backs. When a patient asks him, "What's the success % rate on this surgery?" He replies, "100% or 0% for YOU". It irked me to hear that from him, as an engineer, but he's so damn right when it comes to personal experiences & preparations (emotionally & otherwise)
Not very excited about Linode these days.
SLA numbers are not insurance. At best, they give you back some of your hosting costs... but if you are doing anything serious? that's a very small fraction of what you lost due to the downtime.
Also, 'uptime' means a different thing on a VPS than in a cloud. In a cloud? if a server goes bad, you shoot the server and the customer spins up nodes on another one. the local data is gone. This is not counted as downtime.
On a VPS? that's counted as downtime and as data loss.
Apart from the hacking indecent they have a really good brand with a reputation for doing the right thing. They provide infrastructure as a service, the service is clearly defined and they deliver. Unless they are really sharp then a managed service is not going to be clean, It's full of grey areas and trade off's about whether the team gave correct advice or did the right thing. They should have spun this off to a separate company.
>There is no industry standard for what "managed" means.
Managing expectations here? really hard. Really, really hard. I mean, for $100/month, the provider is probably not going to be involved with the planning and day to day operation; this means you end up with setups where the customer builds a site, gets it working, starts depending on it, and then something horrible happens, and at that point, linode sysadmins will have to step in and fix it. I mean, yeah, a good sysadmin can usually pull it off, but it's "heroic" work, in my mind... You have to sit there and figure out all the weird hacks the last amateur the customer hired used to get the system working. if you ask me to do something like that, I'm going to set an expectation of failure. But that's the problem with managed services, the customer expectation is always success.
This will not be fun for the sysadmin involved. It's much harder to fix a system you are unfamiliar with than a system you are familiar with, and the customer is going to have expectations that the success rate will be as if the sysadmin was familiar with the system. (I mean, hell, I go way out of my way to tell people I sell completely unmanaged stuff, and I still get customers blaming me when their out of date php whatever gets compromised.)
The situations where I would be willing to offer a competing product would be if I could charge (and limit the customers per sysadmin) such that a member of my staff could have weekly or monthly meetings with the customer, going over their architecture and what change have happened, and what problems might happen. (I'm guessing this is going to be more in the $500-$1500/month range, so it's not really competing.)
Alternately, I would be willing to provide a manged service where the customer doesn't have root, except through my tools. Idea being that then all the systems I manage would be substantially similar. I could do this for dramatically less money, had I the time to dedicate to setting it up, and enough customers to make building the tools worth it.
I wish Linode much luck. As a competitor to their unmanaged product, I know I will be referring my customers who need more handholding to this service. There is a whole lot of need in the industry for managed services; VPSs are so cheap these days that people who have no ability or interest in systems administration want to buy them, and they need a lot of help.
My expectation? the customer will generally get a good deal for $100/month. But sometimes? it won't be enough, and that customer will go away very angry and (publicly) disappointed.
Am I missing something here?
There are a number of other services listed on that page, including "Longview Pro - the professional version of our system-level statistics collection and graphing service (currently in beta)."
I got all that from reading the page you linked.
"Up to 8 checks free from Rackspace Cloud Monitoring, for 24x7x365 monitoring of URL content, port, and ping" [1]
"File-level backups. No charge per-server. File storage: $0.10/GB/month. Bandwidth: $0.18/GB/month." [2]
[1] http://www.rackspace.com/cloud/managed_cloud/support_a/ [2] http://www.rackspace.com/cloud/support/
I've also had very few network outages or performance issues in Linode's Dallas datacenter, multiple uptimes of > 1 year on instances and I've only had one unscheduled reboot/failure in almost 7 years of being a customer.
[1] https://blog.linode.com/2013/05/29/introducing-linode-manage...
Fully managed servers are really expensive and often inflexible, while with VPS you are all on your own, which not every developer wants (or feels confident in). I was just discussing a week ago how there is a big market in doing this management.
Hah!
And then sometimes software might need to be restarted which means you have to tell them to restart it themselves or get them to explain enough about how it works so you can restart it.
So unless you are going to charge hourly and staff accordingly, it seems like a no-go.
In this case, $100 per month is really going to pay for maybe two or three hours of sysadmin or application development work max. I.E. helping with various issues that come up in ordinary dev ops or software configuration that are specific to that particular customer's setup. And you just have to count on the idea that most people won't take advantage of more than that average amount of help, like only when they are panicked. And then hope that it is something that you can actually fix in a short amount of time.
If the remote hands are awesome, this is well worth $100 per node. If they are anything but awesome, this wouldn't be worth it for any amount of money.
Their documentation in the Linode Library is also really great -- as a starting point. Assuming that they're using the same guides in recommending server configuration, there are some things that could be done better by a skilled admin. e.g., their LAMP server guide for Debian 6 doesn't include suexec or any variation of FastCGI, two must-haves for a public-facing web server IMO.
For me, the biggest reason for paying for managed services like Heroku etc. is avoiding that risk. This sort of thing would make me far happier to ship some of the load back onto Linode. Can't fault their hardware. :)