How a website should handle a database crash - php

I am currently creating a website in php that has a database backend (can be either MySQL or SQL Server) and I realized recently that if my database crashes at any time, my website will not run properly and probably cause some headaches.
So what is the proper thing to display on the website if my database (or any crucial outside component) goes down? My particular website relies heavily on its database and will be almost useless without it.
One option I have been told is to email the website admin and display a Error 500 page that says something is wrong with the server and basically make the website unusable till the issue is fixed. Is there anything else I could do to work around this problem? Are there any ways to design a website so that the database (any crucial component) crashing isn't an issue?
I am looking for general rules of thumb as well as specific examples of how people have worked around this in the past. Also, these examples don't just have to be for my website example.

If you only have one database server, and the website cannot work without it's database, there is no magic : you'll have to display some sort of nice error page, informing the users there is a technical problem and that the website will come back shortly.
Generally speaking :
Chances of such a problem are pretty low
If your website is a normal one, people will tend to accept a problem once in a while, especially if you communicate about it.
If you can afford it (and have the technical knowledge to set this up), you could use two database servers, with replication (MySQL supports this) between them : one master, which you use, and a slave, that's considered as a backup.
Then, if the master falls, your application will use the slave.
Of course, this will greatly reduce the risks of a database-related problem (having two servers crash at the same time is quite unlikely), but you'll still have problems with all other components -- like your webserver : if you only have one, you might want to consider using two, with the second one as a fallback.
After that, if you still have money (and think you need an even better uptime for your website), you'll want to think about the case when your datacenter has a problem -- setting up server in two separate locations...

The proper thing to display is a simple "oops" error message that gives away no information that would be helpful to hackers. Something along the lines of "We're experiencing technical difficulties" or "website unavailable". This is for security purposes.
It would be good to have an error logging and notification system in place to notify an administrator in case of a crash. That would be fairly simple to write, but I'm sure there are already libraries that handle this. (There's a tutorial with code samples at http://net.tutsplus.com/tutorials/php/404403-website-error-pages-with-php-auto-mailer/ and a simpler example at http://www.w3schools.com/php/php_error.asp)
There are ways to design the architecture of your web site to handle a database component crashing. It's not architecting your website, it's architectin the whole environment. For example, database clustering for high availability (http://en.wikipedia.org/wiki/High-availability_cluster). It's not cheap.
Overall, you just need to ensure that you're doing your error handling properly. A database crash is a classic example ofr why we need error handling. There are plenty of resources and guidance for this.
http://www.google.com/search?q=Error+Handling+Guidelines&rls=com.microsoft:en-us&ie=UTF-8&oe=UTF-8&startIndex=&startPage=1
Edit
I found this and thought it was a very nice resource for answering how to handle the errors:
http://www.nyphp.org/PHundamentals/7_PHP-Error-Handling

It is considered best practice to return a HTTP 500 status code in the event that your database being down, or any other crippled service, prevents your website from functioning properly. Depending on your websites functionality, this could be on a page by page basis or site wide. For example, your "About Us" page may not need database capabilities while your search page would. You could thus keep the "About Us" page up and running but return a 500 status code when someone goes to your search page.
Do not give any technical information about why the site is not working to the end user. This could be a security risk.
If you are using apache, this document will tell you how to setup custom error pages:
http://httpd.apache.org/docs/2.0/custom-error.html
I recommend you use plain HTML for your 500 status code pages. You can also have your PHP pages send a 500 status code via the header() function, documented here:
http://php.net/manual/en/function.header.php

Related

Correct way to handle maintenance mode from a Database setting

building a CMS and looking to find out the correct way of showing a maintenance mode message to users.
My plan is to have the option in my admin backend. I already have the table in my DB for website config and i have a field that just has a value. I was using 1 for normal and 2 for maintenance mode. I just want to be able to update the database.
My question is however, whats the proper way to do this?
My first thought was to check the DB and then if the value is "2" i would redirect the user (all PHP) to e.g. 503.php.
But i have also seen that a better way is in the .htaccess and then i can allow a certain IP to still have access. But can this in the .htaccess be managed from a database check?
Looking to do things properly so curious as the the 'standards' way of doing this.
Is my first suggestion feasible? The main reason for the 'maintenance mode' is actually to protect myself. My site has gone through lots of testing, but i want a way i can quickly shut down access to the site (SEO freindly) just whilst i fix any issues. Or whilst i am updating the site itself.
Thanks
You shouldn't control the maintenance mode setting from the database, if you don't have to.
The database adds another layer of complexity which may fail at some point.
In practice, you have a few options:
Add a text file with the setting for maintenance mode. Prior to serving a request, use PHP to open and read the value of the file. Then, you can handle it appropriately by issuing a redirect if needed.
Use an external service like Pingometer to monitor your site(s). This automates the process and, if any of them go down, use their web hook functionality to send a request to a PHP script that switches this value (perhaps using the method described in #1) to maintenance mode.
I think your goal is that, if happens to the site (externally), your maintenance mode is automatically tripped -- #1 and #2 offer an easy solution.

PHP-based Auth for sickbeard, sab, maraschino

I have a media center which also serves as a low-volume personal nginx server.
Currently, sickbeard, sabnzbd and maraschino are all reached through subdomains, such as sickbeard.domain.com, which are each proxied through nginx to the appropriate port for that service's server. They are each individually secured by their own auth systems, which I don't entirely understand (I tried reading the code, but it's way over my head and in Python, which I know very little about) but they all use the basic auth popup window, which I think is hideous and redundant.
I also have a website, secured by a session-based authorization with a nice form, using php, that I created as part of a tutorial in php (Fort Knox, this ain't.)
What I want is to go to my website, log in to my pretty form, and have links there that take me to all of my services, without having to go through a challenge screen every time. How can I begin to do this? I tend to think my Google-fu is pretty good, but I'm not even sure where to start.
Additional notes:
I put the bones of this together years ago now, but if I recall I went with the subdomain scheme because I was having trouble getting nginx's proxy_pass to work with subfolders. I'm not wedded to it, but I do think it looks nice and clean.
Ideally, I would also like to somehow serve the above services through nginx, so I don't have to have so many open ports.
I also wouldn't mind advice on my php auth scheme. I had a hard time finding tutorials between basic auth and complex systems involving a database of users. I am the only user. I keep my credentials in a flat file outside the path of the site, and I have no need to grow beyond that. I just want an attractive integrated login form, instead of a popup straight out of the 90s.
Sab and Sickbeard are WSGI based, and use the CherryPy libraries. I did a lot of research and decided I could create a new auth method that manually pulled from my php session files and used bcrypt for the hash checks. But I realized I'd stand the risk of my changes being overwritten every time I updated.
Maraschino is also WSGI based, but uses the flasks framework. I had the same realizations as above, but while going through the documentation and code for that, I realized that Maraschino is a lot more powerful than I thought, and the only thing I would want to do on either Sab or Sickbeard that I can't do with Maraschino is non-routine system maintenance, like changing ports or api keys.
So my conclusion is that I'm going to close the ports for Sab and Sickbeard to outside calls, do all my routine activities through Maraschino, and focus my development efforts on getting a better login screen for that. I'll still have multiple ugly auth screens, but I'll encounter them much less frequently. The biggest issue I'll have with that is that when I change my password, I'll have to do it in three different locations.

Partner-website-hosted Web App without exposing source code - Possible?

Apologize if this particular problem has been answered already (a search didn't turn anything directly relevant up).
We are developers of a web app that is used to provide community commenting and "social" to our partners websites. Our app uses Javascript and HTML on the front end, PHP and mySQL on the back.
Currently we are running everything through our own servers, which is getting very expensive.
We would like to ask our partners if we can host the app through their servers, with them getting a discount to our monthly charge due to the bandwidth/cpu load they would help us share.
My question is, is there a way to host our app through our partner's web servers in such a way that we can offload most of the CPU time and bandwidth without exposing our source code?
I would greatly appreciate any ideas/help!!
Thank you very much all!
If you also serve static or rarely changing content your clients could run a caching reverse proxy to remove some load from your servers without giving them any source code at all. But you need to implement caching headers for this to work properly.
You may want to look into nginx.
On second thought: Did you try to compile your scripts using facebooks Hip-Hop for PHP? First of all the script should perform way better, second of all, if you still had to outsource the hosting, you deploy a compiled program, no source code involved.
If you put the code on their server they can find out. So that won't be 100% working. Though you can make it difficult but it's still not great.
Most doable solution might be to separate parts of the application and share them. So: You give away a process (so source and other needed data) but it's only part of the total. That way no partner has your total solution but you do outsource the parts.

Reliable notification system for site errors?

I have a hobby wesbite written in PHP and I like to know if there is a problem with it (database errors, an update broke something, etc.) I have a simple notification system which sends me an email if there is a problem and that would be enough for me. Unfortunately, the mail sending feature of the hosting provider is not very reliable. Usually it works, but there are periods when it simply swallows the mails and doesn't send anything.
Is there some other reliable method for notification of the maintainer in case of an error? It's a hobby site, so I'm looking for something simple. Not an industrial strength solution, but something more reliable than email. How do you monitor your hobby sites?
I tagged the question with PHP, because the site is written in it, but I'm also interested in generic suggestions, not just in concrete PHP solutions.
EDIT: the question is about the mechanism of active notification. I want to be notified when something happens. If PHP email is not reliable then what are the other possibilites of notification?
EDIT2: two examples to illustrate what kind of solutions I'm thinking of:
Store the errors and provide a page listing the latest errors (maybe password protected) which would be polled from my computer which could pop up some window if there is an error. It can work, but it works only if I'm at my home computer.
Use google calendar api to insert an event into it when an error occurs, and google calendar will send me an email reliably. It may work, though it's cumbersome.
some other idea?
Are you looking only for email based alerting systems? If not, you should try Notifo. You can use their API to push notifications and it'll be instantly sent to your phone.
PHP has an error_log function for returning errors in various ways, either via email to an admin, to the servers log file or to an external file. I assume that you could merely substitute this functionality for your mailto when you find an error:
http://php.net/manual/en/function.error-log.php
I've run into the issues you've mentioned with my hobby project as well. When I started I was using GoDaddy who's mail relay was pretty unreliable for delivering mail in a timely fashion.
Two things I'd suggest:
For sending email messages with higher reliablity, check out Postmark. Its a paid solution, but the rates are pretty reasonable and it comes with PHP classes you can hook your code up to fairly easily.
For custom error handling, check out PHP's set_error_handler(). Its a good way to have custom code execute on error conditions on your site. From the documentation:
set_error_handler — Sets a user-defined error handler function.
This function can be used for defining your own way of handling errors during runtime, for example in applications in which you need to do cleanup of data/files when a critical error happens, or when you need to trigger an error under certain conditions (using trigger_error()).
Maybe give Airbrake (formerly Hoptoad) a try. This is a commercial service, but they have a basic free plan (tiny little link at the bottom of the pricing page), and the tool looks pretty cool. It's focused on Ruby on Rails but according to their site has plugins for various other frameworks and languages, inlcuding PHP.
http://airbrakeapp.com/pages/home
We have a system set-up that polls specific pages on our important websites every now and then and checks for certain strings. Would something like that be viable to you?

Deploy PHP website to client server without showing PHP files

I asked a recent question regarding the use of readfile() for remotely executing PHP, but maybe I'd be better off setting out the problem to see if I'm thinking the wrong way about things, so here goes:
I have a PHP website that requires users to login, includes lots of forms, database connections and makes use of $_SESSION variables to keep track of various things
I have a potential client who would like to use the functionality of my website, but on their own server, controlled by them. They would probably want to restyle the website using content and CSS files local to their server, but that's a problem for later
I don't want to show them my PHP code, since that's the value of what I'd be providing.
I had thought to do this with calls to include() from the client's server to mine, which at least keeps variable scope intact, but many sites (and the PHP docs) seem to recommend readfile(), file_get_contents() or similar. Ideally I'd like to have a simple wrapper file on the client's server for each "real" one on my server.
Any suggestions as to how I might accomplish what I need?
Thanks,
ColmF
As suggested, comment posted as an answer & modified a touch
PHP is an interpretive language and as such 'reads' the files and parses them. Yes it can store cached byte code in certain cases but it's not like the higher level languages that compile and work in bytecode. Which means that the php 'compiler' requires your actual source code to work. Check out zend.com/en/products/guard which might do what you want though I believe it means your client has to use the Zend Server.
Failing that sign a contract with the company that includes clauses of not reusing your code / etc etc. That's your best protection in this case. You should also be careful though, if you're using anything under an 'open source' license your entire app may be considered open source and thus this is all moot.
This is not a non-standard practice for many companies. I have produced software I'm particularly proud of and a company wants to use it. As they believe in their own information security for either 'personal' reasons or because they have to comply to a standard such as PCI there are times my application must run in their environments. I have offered my products as 'web services' where they query my servers with data and recieve responses. In that case my source is completely protected as this is no different than any other closed API. In every case I have licensed the copy to the client with provisions that they are not allowed to modify nor distribute it. This is a legal binding contract and completely expected from the clients side of things. Of course there were provisions that I would provide support etc etc but that's neither here nor there.
Short answers:
Legal agreement, likely your best bet from everyone's point of view
Zend guard like product, never used it so I can't vouch for it
Private API but this won't really work for you as the client needs to host it
Good luck!
If they want it wholly contained on their server then your best bet is a legal solution not a technical one.
You license the software to them and you make sure the contract states the intellectual property belongs to you and it cannot be copied/distributed etc without prior permission (obviously you'll need some better legalese than that, but you get the idea).
Rather than remote execution, I suggest you use a PHP source protection system, such as Zend Guard, ionCube or sourceguardian.
http://www.zend.com/en/products/guard/
http://www.ioncube.com/
http://www.sourceguardian.com/
Basically, you're looking for a way to proxy your application out to a remote server (i.e.: your clients). To use something like readfile() on the client's site is fine, but you're still going to need multiple scripts on their end. Basically, readfile scrapes what's available at a particular file path or URL and pipes it to the end user. So if I were to do readfile('google.com'), it would output the source code for Google's homepage.
Assuming you don't just want to have a dummy form on your clients' sites, you're going to need to have some code hanging out on their end. The code is going to have to intercept the form submissions (so you'll need a URL parameter on the page you're scraping with readfile to tell your code that the form submission URL is your client's site and not your own). This page (the form submission handler page) will need to make calls back to your own site. Think something like this:
readfile("https://your.site/whatever?{$_SERVER['QUERY_STRING']}");
Your site is then going to process the response and then pass everything back to your clients' sites.
Hopefully I've gotten you on the right path. Let me know if I was unclear; I realize this is a lot of info.
I think you're going to have a hard time with this unless you want some kind of funny wrapper that does curl type requests to your server. Especially when it comes to handling things like sessions and cookies.
Are you sure a PHP obfuscator wouldn't be sufficient for what you are doing?
Instead of hosting it yourself, why not do what most php applications do and simply distribute the program to your client with an auto-update feature? Hosting it yourself is complicated, from management of websites to who is paying for the hosting.
If you don't want it to be distributed, then find a pre-written license that allows you to do this. If you can't find one then it's time to talk to a lawyer.
You can't stop them from seeing your code. You can make it very hard for them to understand your code, which is a good second best. See our SD PHP Obfuscator for a tool that will scramble the identifiers and the whitespacing in the code, making it much more difficult to understand.

Categories