Hacker News Reader

Comment by dpcan | Hacker News Reader

I'm not sure why you are using prgmr over Linode, Slicehost, or the Rackspace Cloud servers. The others have really simple backup tools.

View on HN

Because prgmr is cheap and I wanted to keep costs under control from day one. :)

For example, there's no point in me paying for a nice control panel when I can set up Apache or nginx in no time on my own. I'm not a super sysadmin, but I know enough to quickly write a script that backs up code and database to Dropbox at regular intervals and so on. So paying for those extras? Not sure I see the point when you're not even covering hosting costs :)

Friendly reminder: test restoring from that, regularly. After your hard drive crashes is a bad time to learn that, e.g., zipping up a folder of MySQL files is not exactly backup best practice.

Thanks, great advice :) Right now I'm taking backups of dumps - slow, but a tiny bit easier to deal with. At some point I'm thinking about replicating to a slave somewhere, although to be honest I need to do some more reading about best practises.

Yes. Obviously, specifics will be dependent upon your exact situation, but it would be nice to have a small road-map for backups. I occasionally take my database dumps and fire up postgres on the home server and make sure I can restore the database. That's about as far as I take it usually. Probably not a robust failsafe, but it's what I do now.

I'd expand this to actually take the time to implement (or make yourself) a system to automate said backups and maybe even testing them automagically. You'll probably end up using the same set of scripts for all your beginning projects and it'll be worth having.

Don't forget to check logs to make sure your automation actually works.

Periodically make full restore from bare metal or fresh VPS image using your backups and following written procedure.

Make corrections if something does not work as expected.

if you don't mind me asking, what package are you using on prgmr.com and the level of traffic it handles.

I've been looking at some VPS options and I wanted to make sure I select the right one that is not oversold + performs well.

http://journal.uggedal.com/vps-performance-comparison/

Thanks. That is a really useful article. I may switch from slicehost to linode as a result, especially as my 1024MB slicehost slice has struggled badly under highish (30,000+ uniques a day) traffic.

I may switch from slicehost to linode as a result

I'd heard rumblings in the past to the effect of "Linode has better performance than Slicehost," but I hadn't really seen such a detailed, quantified comparison until now. Now I'm thinking about switching as well.

My Slicehost slices have never impressed me from a performance standpoint, that's for sure. Their admin console and service have kept me in the fold so far, but I think it's time to at least spin up one Linode instance and do some re-evaluating.

if you don't mind me asking, what package are you using on prgmr.com and the level of traffic it handles.

I'm using the 1GB offer. This is something that I may blog about briefly, so far it's running TweetingMachine and a couple of other tools, and has supported the Hacker News traffic decently as well. That's most likely thanks to me using nginx and w3 total cache on the Wordpress blog. That said, if someone really wanted to DDOS me, it wouldn't take much for it to fall over ;)

How about a write-up on how to get prgmr up and running? If it doesn't take a super sysadmin, it might be nice to see the process you took.

How about a write-up on how to get prgmr up and running?

I could do, but honestly it doesn't differ much more a standard Linux setup - the only difference is a lack of control panel.

I think you're drastically underestimating the cost of a crash and the need for a good backup.

If you go down, how many customers may you lose for life?

What do you tell those users if you happen to lose some of their scheduled tweets?

How many refunds will be requested?

Not to mention the time it will take you to bring everything back to life on a new server.

I just looked at prgmr (I hate that name), and it looks like you can get a really really tiny VPS for $48 for a year. That's silly cheap.

I'm teaching a couple of the users on thingist python, and I've been wanting to give them accounts on a box so that they could have a real-world sandbox to play in...$48/yr lets me afford that pretty freaking easily.

prgmr sounds freakishly awesome. What was that site the other day? showoff.io or something? $5/mo for a VPS vs $5/mo for a dynamic DNS updater/NAT traversal mechanism.

(Just waiting for my instance to spin up now)

An Amazon EC2 Micro instance is pretty close to that price, plus you don't have to pay for it while it's not running, and you get to learn AWS too.

An amazon EC2 micro instance is something like $60/mo. I know this because I tried to sign up for their "free" tier a few months ago, and got charged. Luckily amazon had my account information on file, so they went ahead and just charged me for the balance.

(It took a few weeks/months to get my money refunded...and forever convinced me that amazon's customer support is completely worthless)

I'm sure AWS works for some people, but that left a pretty bad taste in my mouth.

I feel as though you may have encountered the same problem as I did when starting my Amazon service.

The micro instance - the free tier - is actually free, and if you run more than one, additional ones cost roughly $15/month.

However, what most likely happened is that when spinning up your first instance, you started a "Small" instance instead of a micro. "Small" is the default server selected when adding a new instance through the AWS webpage, and I did the same thing without realizing right away that it was not the free tier.

Small instances cost roughly $67.50 per running month, so it would seem to me that this could have been what happened to you.

Yeah, I think that's what happened to me to.

I guess it just felt kindof "bait-and-switch"y to me. Imagine going into a restaurant that says "free buffet", where there are some non-free things mixed in that aren't really labeled all that well.

The thing is, yeah, I made a mistake and clicked the wrong thing. That happens. What was frustrating to me, though, and why I'll never use AWS again was how long it took to find out what happened. There was no support email, no support chat, nothing.

> An amazon EC2 micro instance is something like $60/mo.

An EC2 micro instance is around $15/mo (0.02 x 24 x 30). If you were charged $60 (0.085 x 24 x 30), you were running a small instance. They aren't the same thing, and Amazon did/is doing a shit job of making people looking for the freebie account aware of the difference.

AWS is pretty awesome if it meets your needs. We use it at work, and don't have many problems with customer support. That said, it seems like there's obvious support prioritization based on account value.

I'm sorry you had a bad experience. For what it's worth what you saw was more terrible marketing/support handling for freebie accounts than anything related to the service as normally used.

Amazon's pricing is complicated for this kind of price comparison - it actually costs ~$150/yr to run a micro instance full-time.

Here's the math: http://news.ycombinator.com/item?id=2049196

Is this the primary problem you see with my service? you think this is a bigger problem, for instance, than my provisioning delays? (e.g. it usually takes a day or so for a new customer to get set up.)

I'm actually working on a cheap low-performance storage service designed for backups (webdav and zfs snapshot accessible. I'm targeting 1/10th the price of s3 before bulk discounts, but I also am also targeting a lower level of redundancy and performance.) but I thought the primary demand would be people on other services or on their own servers wanting a backup, so it's interesting to hear if other people place a much higher priority on ease of setting up backups.

Because if you know your unix and you have the time and experience to quickly setup the stack you need there's no need for a pretty control panel. If I only need ONE VM for something I know it's trivial to setup for me and will probably very seldom require me to use a admin console, I'd always go with pgrmr. There are situations though, where going with a bigger and shinier provider might make sense, but most of the time pgrmr is mighty fine for a beginning hobby business.

Reply on news.ycombinator.com