Reticularium - Servers

Rackspace Cloud Servers outbound bandwidth limits

2012-01-25T13:45:00+01:00

Is there a throughput limit on my server’s network interface card?

Your Cloud Server has a bandwidth cap that is applied in relation to the size of the server. Please note that the bandwidth caps are placed on outbound bandwidth only. Inbound bandwidth is not capped.

Server Size Public Limit ServiceNet Limit

256MB	10 Mbps	20 Mbps
512MB	20 Mbps	40 Mbps
1024MB	30 Mbps	60 Mbps
2048MB	60 Mbps	120 Mbps
4096MB	100 Mbps	200 Mbps
8192MB	150 Mbps	300 Mbps
15872MB	200 Mbps	400 Mbps
30720MB	300 Mbps	600 Mbps

Chart valid as of December 8, 2011

VPS network bandwidth limit

2011-08-27T17:42:00+02:00

This doesn’t harm in most cases, but still is worth keeping in mind that if you have a 256MB VPS, it is a 1/128 of the 32 GB server and thus its network bandwidth is usually limited to 1/128 of 1 Gigabit which equals 1 MB/sec.

Various gateway servers, balancers, caching proxies don’t need CPU power and RAM, so small 256MB VPS choice looks natural. This is where the problem may arise.

M-Script's new home

2011-07-31T11:16:00+02:00

M-Script project has been moved to its new home: http://m-script.org

It’s been under heavy development lately: old scripts have been renewed and enhanced, folders reorganized, files moved to proper locations, default monitors made more informative.

It’s also been published on Github

LOAD DATA INFILE vs bulk inserts

2011-05-13T08:58:00+02:00

This and a bit more here

With graphs!

Everybody knows that LOAD DATA INFILE is fast, many did benchmarks, but nobody cared to publish results with graphs, so that others could use them to illustrate the approach :) Thanks a lot, Kevin

Amazon S3 as a static website hosting

2011-02-18T10:51:00+01:00

Amazon announced that S3 buckets now can be used for hosting static websites, not just files and images. It’s achieved by adding two properties to S3 buckets: document root and custom error pages.

Looks pretty useful. It was possible before, either by mounting S3 bucket as a folder via S3FS or using complex rewrite rules, but it’s become so much easier now.

For example, this article, if you are reading it from the website, is a static HTML page saved to disk by RoR cache controller. Using RoR routing I can easily remap such pages to e.g. static.reticularium.com and place this whole sub-domain to S3. This would decrease the load on my server and make access to such pages faster (via CDN). Not sure I want it though, I have to double-check whether it affects SEO things, but at least it’s possible now. For one project or another, it can be used to make access to your content faster and probably save you some money.

S3 file size limit raised

2010-12-20T09:46:00+01:00

Files stored on Amazon S3 can be 5TB in size now.

If you are using my backup scripts to backup your data to S3, you can now empty the split_size parameter.

Cloud computing definition

2010-12-11T19:17:00+01:00

Despite the title, I am not going to provide any definition. You can find many definitions made by much smarter guys on the Internet. I saw many, but after reading most of them I had a stronge urge to say “Wow, it’s fantastic! So what you say cloud computing is?”

In fact, there is no definition. Because cloud computing is a paradigm shift. If you still need a definition, take this as one. I mean “A paradigm shift”. Shift towards what? Well, nobody knows exactly.

That’s why I am going to just explain what is usually meant by this, and I’ll try to do it in a sorted way.

In fact, there is no definition. Because cloud computing is a paradigm shift. If you still need a definition, take this as one. I mean “A paradigm shift”. Shift to what? Well, nobody knows exactly. What we know is direction.

That’s why I am going to just explain what is usually meant by this, and I’ll try to do it in a sorted way.

To understand the direction of abovementioned “paradigm shift” it is necessary to start from the point where we were before that cloud computing buzzword appeared. So here is the traditional stack:

Infrastructure	Hardware: servers and network
Host operating system	These two are optional and can be replaced by a single row representing the more traditional “one server – one system” approach. I wanted to emphasize here that virtualization alone doesn’t mean cloud computing.
Guest operating system
Platform	LAMP, for example. Includes frameworks, e.g. Ruby on Rails, Symphony, Django..
Software	Your own application or CMS, blog engine e.g. Wordpress and so on
Content	Texts, images, videos or whatever users can see

In spite of early experiments on cutting the fat out, the stack stays the same in cloud environment. So each of its levels can be put into the cloud, and this is the main reason why it is so difficult to make a single definition. This is why you might find cloud computing definitions that mean rather different things: they are about the different stack levels. By the way, there are those ugly acronyms like SaaS (Software as a Service), PaaS (Platform as a Service) and so on. Think of the “aaS” part as “in the cloud” and you can get all of them from the stack above, like this: software in the cloud = SaaS, platform in the cloud = PaaS, Infrastructure in the cloud = IaaS.

So let’s put the whole stack “into the cloud” to see what has changed:

Infrastructure	In the cloud everything is virtualized. So servers are virtual, as well as storage, network interfaces and so on. There is hardware of course, somewhere behind, but you don’t need to know what hardware there is, neither you want to know where it is located (except from connectivity point of view) and how all these components are connected. The most pertinent example is AWS of course
Host operating system	These two don’t exist in the cloud (yet). See my speculations below.
Guest operating system
Platform	It’s not a cloud server with some platform preinstalled. It’s easy to distinguish: traditional platform is something you buy with some server (or part of a server) and certain operating system. But if you buy a cloud platform, you never ever should bother about operating systems and anything underlying at all. You only need to know how to interact with the platform. That is, how to upload the code and deploy the app. The best example here yet is Google AppEngine
Software	Basically, everything that is able to run in your browser. Note that it should run, not be shown. I think the most relevant example is Google Docs. Note that similar to platform we don’t even need to know about anything related to the previous stack level, that is, about platform in this case.
Content	Texts, images, videos or whatever users can see. Looks the same, huh? It’s because this stack level was always in the cloud. You never needed to know where text or images were located or what software, platform and operating system were installed on the server sending those files to your browser. This is the point. Since you don’t know the details, you can’t draw them on diagrams, and you draw something like a cloud instead. This is where the term “cloud” has appeared from.

So, now, what I think a cloud operating system¹ could be.

Theorizing based on tables above, I’d figure it as:

something that the uplying stack level (Platform) is installed upon.
something making unnecessary for Platform to “know” about anything below its level, i.e. certain cloud environment, number of servers, OS they have, their location and so on.

Does it sound fantastic? Not at all. Take a look at Eucalyptus and Openstack. Basically, they are cloud management systems, and I am not sure they will turn out to be a full-featured Cloud Operating System some time in the future. But they are definitely a step in this direction.

¹ Don’t confuse with cloud-oriented operating systems, either web-based or traditional. Despite the fact that they are often referred to as cloud operating system, too, it is not correct and they are not what I mean.

MySQL export to CSV

2010-02-02T22:12:00+01:00

Pretty easy:

mysql -u<username> -p<password> <dbname> -B -e "select * from <tablename>" | sed 's/\t/","/g;s/^/"/;s/$/"/;s/\n//g' > <filename.csv>

auto-whitelist: Inappropriate ioctl for device

2010-01-04T20:59:00+01:00

If Amavis reports this scary error, it doesn’t necessarily means that your disk is cracked. Usually this means that the file it is talking about has wrong format. This happens, for example, if you upgraded Amavis or Spamassassin. Check current format with the file utility and upgrade it with dbX.X_upgrade where X.X is appropriate version. Or simply remove the auto-whitelist file, it will be re-created. Same with bayes_seen and bayes_toks files.

Rackspace confused by VAT - part 2

2009-11-28T01:56:00+01:00

I have just found where that Rackspace decision to add VAT to the price resulted from. First I wanted to edit the previous article, but then decided to leave it as is. It’s still correct except a couple of details.

These details are as follows.

There are new EU VAT regulations forthcoming. The best explanation I’ve found is the so-called Notice 741A from UK HMRC website.

The most important moments there are:

“From 1 January 2010 there are two general rules for the place of supply of services, one for business to business (B2B) and one for business to consumer (B2C) supplies.”

“The B2B general rule for supplies of services is that the supply is made where the customer belongs”

“The B2C general rule for supplies of services is that the supply is made where the supplier belongs”

Thus, starting from 2010, Rackspace will have to pay VAT if they supply their services to non-business customer (private individual or any body with no business activities or any body who receives a supply wholly for private purposes).

So, I agree that Rackspace did have a good reason to add VAT to the price for non-business customers. And the VAT they add should be the UK VAT, this is correct too in accordance to the new regulations.

But still two important questions left:

Notice: I am talking about Rackspace only because they were the first (as I know) who were kind enough to warn about these changes. I am afraid there going to be others.

The first question is how Rackspace is going to tell business from non-business customers. Notice 741A explains possible options:

“VAT registration numbers are the best evidence that the supply is not received for a wholly private purpose and should be requested. If your customer is unable to provide a VAT number, you can accept alternative evidence. This includes certificates from fiscal authorities, business letterheads or other commercial documents indicating the nature of the customer’s activities. Such evidence should be kept as part of your records. Where VAT numbers are available, they should be shown on your invoice.”

Rackspace in their message and on their website mentions VAT number only. This affects sole traders and small businesses that are not registered for VAT. I am sure there are plenty of such customers. This affects me too since I operate as a sole proprietorship and I am not registered for VAT. So I will have to get a clear answer from Rackspace regarding this, follow the updates.

The second question is more complicated. Why Rackspace is going to add VAT instead of deducting it? Let me explain.

There are 7 possible Rackspace/customer relationship combinations. Look at the table:

Rackspace location Customer type Customer location Who pays VAT now Who will pay VAT in 2010

UK	Business	UK	Customer	Customer
UK	Non-business	UK	Rackspace	Rackspace
UK	Business	EC / non-UK	Customer	Customer
UK	Non-business	EC / non-UK	Customer	Rackspace
UK	Business	non-EC	Customer	Rackspace¹
UK	Non-business	non-EC	Customer	Rackspace¹
US	Any	Any	Not Applicable	Not Applicable

¹ I am not sure (I assume nobody is), the rule of “where it is used and enjoyed” applies here, and telecommunications services is a pretty specific case. I think that I am correct here, because e.g. TV broadcasting is used and enjoyed where customer is located but servers hosting logically is rather used and enjoyed where datacenter or supplier is located.

I would assume that non-EC customers will be supplied from US office, so it’s not the case. But this may be the case if related to other service providers established in Europe and their non-EC customers.

One obvious moment is notable: UK non-business customers that are currently paying prices without VAT will be paying price + VAT despite of new regulations don’t affect them!

And finally this comes to the main question: why VAT should be added to the prices? My vision is that prices should remain the same except prices for business customers, where VAT should be deducted—because business customers pay it themselves (you may want to google about the “reverse charge” or “tax shift” procedure).

I know that there is no VAT in the USA, so it looks kind of logical to add it to the price that USA customers pay. This fact is always confusing European customers and they agree to pay extra money. But it’s wrong. I have USA clients and I don’t have corporate tax. Should I add US corporate tax to my usual prices for my USA clients?

OK, another calculation. According to Wikipedia, in the US the average combined federal and state corporate tax rate is estimated at 39.3%. In the UK corporate tax is 21 to 28% and VAT is 17.5% which totals to 38.5 to 45.5%. I know that this is quite inaccurate, but accurate enough to give an idea: operating from Europe, US companies have less tax burden if they claim that their usual prices don’t include VAT. They like this situation of course, so, when they are forced to pay VAT, they want to add it to the price to maintain status quo.

As I mentioned on Slicehost forum, it’s not about the price. I inderstand, there may be higher business and operational expenses in Europe, so if they simply had separate prices for USA and Europe, I would understand. But I feel like me and all European customers are being fooled with that VAT trick. Correct me if I am wrong..

Rackspace confused by VAT

2009-11-27T13:38:00+01:00

Yesterday I received an email message from Rackspace:

“Dear Igor Simonov,

The Rackspace Cloud would like to inform you of an important change to our billing system that will enable us to correctly handle VAT invoicing for our European customers.

As an existing Rackspace Cloud customer, you will need to enter your VAT registration number in your control panel prior to January 1, 2010 to avoid being charged the 17.5% tax for VAT. If you do not have a VAT ID number, or choose not to supply one, The Rackspace Cloud will begin charging you the 17.5% VAT tax on all invoices after January 1, 2010.”

There is an update for this article

Yesterday I received an email message from Rackspace:

“Dear Igor Simonov,

The Rackspace Cloud would like to inform you of an important change to our billing system that will enable us to correctly handle VAT invoicing for our European customers.

I got stuck for a little. From what I know about VAT nothing can result in any price difference for customers are they VAT payers or not. So I have come up to conclusion that Rackspace billing guys just don’t understand what VAT is and how it works. They might have been confused by necessity to enter it in invoices as a separate line, and they obviously missed the point: if I am not a VAT payer, it’s me who don’t include VAT in my invoices, and it’s my clients who pay VAT instead of me. It’s me who is a client for Rackspace, not vice versa.

Here’s a simplified example:

Rackspace produces a service from scratch and sells it to me for 10 dollars.
I produce my service based on a service provided by Rackspace and sell it for 20 dollars.
My client offers a service based on my service for 30 dollars.

If there is no VAT at all (e.g. if these three entities all are in USA), everything is clear. Each participant makes 10 dollars and pays taxes calculated on this basis. Income tax, corporate tax, whatever.

If there is VAT and each side is registered as a VAT payer, it looks like this:

Rackspace supplies their service to me for 10 dollars and pays VAT tax based on 10 dollars
I receive their service and supply my service for 20 dollars. From their invoice I get their VAT and claim that the service I’ve received had VAT included in its price, so I pay VAT based only on my 10 dollars earnings, i.e. same amount as Rackspace.
My client supplies his service for 30 dollars, but like me he pays only VAT based on 10 dollars, because his cost basis has already VAT included.

Total VAT calculated on basis of the final service price is exactly the same as the sum of VATs paid by all participants of this chain together, but each pays the own part.

Well, this is simple. Now where has Rackspace mistaken.

If Rackspace is a VAT payer, I am not a VAT payer and my client is a VAT payer, how it should be:

Like in the previous case, Rackspace supplies their service to me for 10 dollars and pays VAT tax based on 10 dollars
I am not a VAT payer and therefore I cannot deduct Rackspace’s VAT from my VAT, but I don’t need this – I am not a VAT payer! I still should receive the same service from Rackspace for the same price, the only difference is that I don’t care if they enter their VAT in invoice or not.
Now, if my client receive my service for the same 10 dollars price, it’s no good: I cannot show VAT in my invoices, so my client will pay VAT based on the whole price, i.e. on basis of 30 dollars. Apparently this cannot affect Rackspace at all, they already have gotten their money. For me personally being or not a VAT payer doesn’t make much difference, because:

my services are not being re-sold
my prices are lower by (at least) VAT percentage – because I don’t have to pay it.
95% of my clients are from USA anyway

Note: to make for my clients possible to pay VAT I don’t supply my service for 20 dollars. My actual price is 10 + 10 / 117.5 * 100 = 18.51, so my clients still can supply their service for 30 dollars and make same money as if I was a VAT payer.

How it looks like if Rackspace doesn’t change their decision (I hope they do):

They simply supply their service for higher price to non-corporate users. And they will have to pay VAT based on 10 dollars plus VAT based on additional 17.5% dollars that they call VAT but in fact it is just a higher price.
I can’t supply my service for $18.51 anymore, I want to keep my earnings unchanged, so my clients can’t supply their services for $30 for the same reason, they have to pay my VAT too, and I can’t compensate it, because the compensation is paid to Rackspace, but it should be paid to the treasury! This is very very wrong.

So in this scenario I make less money, my clients make less money, treasury gets same taxes, Rackspace is the only one who gets more money. Well, probably this is the real reason..

To make this even more idiotic I have to mention that my VAT, if I paid it (Czech Republic) would be 19%, not 17.5% :)

Rackspace Cloud at the first glance

2009-11-22T23:47:00+01:00

Rackspace Cloud is a division of Rackspace focused on cloud technologies development. Their own service (actually three – Rackspace Cloud Servers, Rackspace Cloud Files and Rackspace Cloud Sites) is based on the recently acquired Slicehost VPS servers infrastructure. Since I am a Slicehost client myself and have many clients being hosted there, I was really curious how the Rackspace services look like.

Because of my relocation to Prague I was too busy and didn’t have a chance, but better later.. so starting now.

Creating an account was a bit disappointing: when I tried to log in to a just created account, it didn’t allow claiming that my account is suspended. Their live support (they call it “our famous Fanatical support”) explained me that it is not a bug, it is a feature, needed for fraud protection, and my account will be activated only after they have me verified by a phone call. I didn’t understand if it’s a random procedure or every new client has to be verified, neither do I know how long would I have to wait if I didn’t address the question to their live support right away. Also, I would feel much more comfortable if I got some warning that this may or should happen.. But OK, no real problem, they called me in several minutes and the rest of the process was absolutely seamless.

They have a really cool interface, well designed, pretty looking and quite intuitive. I got my server up and running in minutes and found that it looks exactly like my Slicehost servers. Two IPs – internal and external, initial root password which is a combination of the server name and random part and so on, everything looks familiar. Quite expected. But there is one notable difference. I don’t know why, but Rackspace 256M server shows 329036k of RAM while Slicehost’s 256M slice shows 262364k as expected. I don’t know if there is similar difference between higher memory slices, I’ll check this later. Disk space is same – 10GB for 256M server, 20GB for 512M and so on.

Update: Looks like in my case Xen balloon driver doesn’t work properly. According to /proc/xen/balloon 262144kB is allocated, 270336kB is the maximal target, but I see 329036kB of usable memory. Also, there is no usual message about “Setting mem allocation” in dmesg. But shhhh.. I don’t mind having a bit more RAM :)

I am going to play with both APIs a little and describe this in another article, but in general both APIs are similar (yes, Slicehost and Rackspace have different APIs).

So, now it’s time to look at pricing.

Slicehost offers 256MB slices for $20 a month, and it looks more expensive than $10.95 a month from Rackspace. But don’t forget about bandwidth. Slicehost’s 256MB plan includes 100GB of bandwidth, similar plan from Rackspace doesn’t include any bandwidth. They count it separately: $0.08/GB for incoming and $0.22/GB for outgoing bandwidth. Thus, if incoming to outgoing ratio is e.g. 1:10, you will pay the same $20 a month for 256MB server and 40GB of outgoing bandwidth as you pay for Slicehost’s 256MB slice and 100GB of bandwidth. So, it depends. If you don’t need more than 40GB/month for a 256MB server, Rackspace is cheaper for you. On the other hand, Rackspace charges on the per hour basis, you can turn servers on and off via their web interface or use API and pay only for time when servers are on.

This comes to another interesting question: Slicehost and Rackspace servers share the same internal network. I am pretty sure that they treat any data transfer between them as local and don’t charge for it. I am going to check this tomorrow (billing statements available for previous day only). If it’s true, it’s possible to use Slicehost’s server as a frontend/proxy and save some cost.. Updates will follow.

Java VM initialization error

2009-10-10T11:21:00+02:00

Error message:

12	Error occurred during initialization of VMCould not reserve enough space for object heap

means: not enough Virtual Memory for Java VM to initialize. Weird for a first glance, but are you sure you have it enough? Quite probable that you have ulimits setup, which, in turn, is quite probable if you have cPanel.

Injected iframes

2009-10-06T20:14:00+02:00

Looking through job alerts I receive by email from oDesk I often see requests similar to this:

Some malicious user infected all htm and php files off our server with a iframe in the end off each file. I need to remove it from all files

These requests make me think that some webmasters don’t understand how it works.

Looking through job alerts¹ I receive by email from oDesk I often see requests similar to this:

Some malicious user infected all htm and php files off our server with a iframe in the end off each file. I need to remove it from all files

These requests make me think that some webmasters don’t understand how it works.

This wasn’t some malicious user who infected your files. This was you!

So, how it works (at least how it worked in all the cases I had to deal with).

You or your developers use FTP to upload edited files. Some of the computers connecting to the server via FTP have Windows installed, which is infected by a trojan. That trojan discovers FTP password and makes it accessible for botnet. Botnet is a network of other infected desktops controlled by, well, some malicious person, but not your user! That guy probably never seen your website at all. Each of the computers in that botnet now can download any file from your server, add iframe to it and then upload it back again. They do it again and again, not only because you probably remove those iframes, they also have to change links in those iframes quite often. So, if you simply remove those iframes, it won’t help.

What you can do.

I am not going to insist that you or developers should use different operating systems (some of my clients forced developers to do this though). I personally don’t believe that there exists such thing as non-infected Windows computer connected to the Internet in this world, but many people do. So, if you believe, you may rely on antivirus software and try to find and destroy the trojan. Then change FTP passwords, remove injected iframes from your website and wait if they appear again.

Better approach is to disable FTP access to your server and use SFTP or SCP to upload files. Maybe trojans will learn to recognize them too, but I haven’t seen such trojans so far. If you can’t or don’t want to use SFTP or SCP and disable FTP, there is another good idea: you may block access to FTP for all IPs except those you need access from. This will block access from botnet computers but trojan still can use developers’ computers to access your website.

Of course, you may use all three recipes simultaneously together with some script searching for and removing injected iframes every 5 minutes, and this combined solution is probably the best you can do.

¹ Note for my current employer: I am not searching for a job, I am just curious enough to take a look sometimes :)

Another way to make changes on all tables in PostgreSQL

2009-09-27T20:42:00+02:00

For example, to grant all privileges on all tables of the current database to user username:

select 'grant all on '||schemaname||'.'||tablename||' to username;' from pg_tables where schemaname='public';

Change username to what you need.

The output of this is a list of commands that that can be copied via clipboard and run all at once.