Installing Ansible on a fresh Mavericks system.

I recently purchased a new Retina Macbook Pro, and was surprised that Ansible did not install as easily as it did before. Looking back through the output on the CLI we see it is trying to use Apples CLANG. Which unfortunately has some deprecated stuff. I guess i had already installed the correct GCC etc manually on my older machines.

No worries though, this is the simplest way i have found to power through this.

Firstly, install apples developer tools. Likely you will just download XCode through the App Store. Then once you have opened it at least once to agree to the terms and conditions, you will also need to go to the xcode preferences and ‘Locations’ tab, to select the latest ‘Command Line Tools’ version. Then you will need to run this on the command line.

This will install some additional command line tools which you will need. It will open a GUI prompt to install them, so just go ahead with that.

With XCode installed we can now install Homebrew, a package manager for mac, which will make installing the non Apple version of GCC much easier. To install Homebrew, run the following on the CLI.

Then run this, which will tell us if Homebrew is working correctly, or if there are any issues.

Now we must install the correct GCC via Homebrew.

With that now installed, we must move Apples GCC aside so that we can use the correct one.

Then, we are good to go ahead and run the steps on Ansibles installation page found here

Problems that were occurring previously with CLANG and the like should now be gone.

Credit where credit is due, these 2 sites documented the steps i followed:
[1] Installing Homebrew on OS X Mavericks 10.9, Package Manager for Unix Apps
[2] Paul Crawford : Installing gcc on Mavericks

Symfony2 and “ContextErrorException: Warning: SessionHandler::read()” Exception

I just upgraded to Symfony 2.4, and started randomly getting this error:

ContextErrorException: Warning: SessionHandler::read(): open(/var/lib/php/session/sess_3cq63tff1lhbghus3o9jsh26j4, O_RDWR) failed: No such file or directory (2) in /var/www/codeconsortium.com/public/vendor/symfony/symfony/src/Symfony/Component/HttpFoundation/Session/Storage/Proxy/SessionHandlerProxy.php line 69

Which was pretty odd, as there is a discussion about this, which you can find here on symfony/symfony github, However the issue seems to be in part with PHP, but it was for me already patched. As the issue is introduced in 5.4 somewhere and fixed in 5.4.11.

It seems that everything was fixed, and i was many versions ahead. So why was this cropping up randomly now? I had not changed my PHP version or my OS version, just symfony.

It seems some config changed, which i show below. If you do not have this set, then i believe it falls back on the default php.ini setting of session.cookie_path, which if blank causes this. Add this config below will fix the issue.

I hope this helps.

Stylish Pure CSS 3 Bread Crumb Bar.

I was wanting a new bread crumb bar for my site CodeConsortium and i spent considerable time looking for a pre made solution that did the following:

  1. Used no images at all including background images
  2. Tag shaped and overlapped like in the google style on their docs pages
  3. Used very lean and semantically correct HTML

After failing to find anything on google, i decided i would need to write my own and so i did just that. One of the things i found was that most of the bread crumbs i found either used background images, or used psuedo-selectors to fake the triangular end which unfortunately was being done using the border trick instead of the rotation trick. Those using the rotation however were bland as they could not use images, and did not use gradients to make it all stylish.

The solution i came up with is to use the psuedo div (:before and :after) at a -45 degree angle and use a proportionally rotated gradient to match that of the main crumb. It was very tricky but here is what i came up with looks like:

Screen Shot 2013-02-27 at 04.43.23

and when you hover over a section:

Screen Shot 2013-02-27 at 04.43.48

Not bad huh? The last section is indented to convey that it is the current milestone along the path of bread crumbs.

After deciding that the best balance between lean html and a semantically correct bread crumb navigation would be to use just an ordered list inside a nav element, i came up with the following html:

Thats fairly lean, and still reasonably ‘semantically correct’. Now to add the CSS to style this puppy:

Oh one note, i have not tested this yet in IE (sorry), but if you do find an issue in IE, or you can improve upon this in some small way, let me know and ill fix it.

Nikto Server Auditing and Resolving Issues.

If you are security conscious and want to find an easy way to determine what aspects of your server setup are presently vulnerable to known exploits; then you may want to try a server security auditor/scanner. There are lots of security scanning/auditing scripts and apps out there, including some websites that will audit your site and provide you with a free report.

Here we will look at the basic usage of Nikto 2 and some of the common issues that it points out, and how we can resolve them. This guide is targeted at users running a LAMP stack (Linux, Apache, MySQL and PHP). However, it may still apply to some other setups out there. Hopefully the information in this article will help get you started.

If you do not have Nikto already, you can download it here.

Usage is relatively simple, just type:

$ perl nikto.pl -h yourdomain.com

Or do a more comprehensive scan with:

$ perl nikto.pl -j yourdomain.com -C all

Here are some common results you will get and what some things to consider when remedying the issue.

+ Cookie PHPSESSID created without the httponly flag

+ Cookie __cfduid created without the httponly flag

The __cfduid cookie is set by CloudFlare, so you won’t see this if you do not use CloudFlare. Its nothing to be concerned about.

+ The anti-clickjacking X-Frame-Options header is not present.

How to set the X-Frame-Options Response Header. Just add this code snippet below to your /etc/httpd/conf/httpd.conf file. Don’t forget to restart the httpd server.

After setting this, and running another test, you might find you now get this:

+ Uncommon header ‘x-frame-options’ found, with contents: DENY

I am not sure why this is, my best guess is that nikto expects SAMEORIGIN instead of DENY. Either of which is fine though. Unless you know otherwise, i would just ignore this at this point.

+ Uncommon header ‘cf-cache-status’ found, with contents: MISS

This is CloudFlares cache for your sites assets. Make sure your servers clock time and your httpd.conf’s headers are set properly to ensure that the sites assets are not interpreted as being stale. You may need to login to CloudFlare, purge the cache and visit the site a few times to ensure the CloudFlare cache is up to date. Then this issue should be resolved.

+ Server leaks inodes via ETags, header found with file /n8YeaczG.pl, fields: 0x3c3 0x4bbd982a52140

Disable ETags from within your /etc/httpd/conf/httpd.conf by setting the following:

+ robots.txt retrieved but it does not contain any ‘disallow’ entries (which is odd).
+ “robots.txt” contains 7 entries which should be manually viewed.

If you have a robots.txt file, you may receive one of the 2 above messages, they are not that important. If it says you do not have any disallow entries, while it is not much of a security risk (unless you are exposing sensitive data in html files [which is ridiculous and you shouldn’t be doing that]) you should add some default disallow entries. The reason why this is, search engines will use this file to determine what pages on your site should be indexed and cached etc. By adding some disallow entries for pages that will present just forms or stuff that is useless to your users google and other search engines can focus on indexing the pages on your site that really count. This is important because since googles panda update, pages on your site that have a high html to content ratio are actually really bad for your sites page rank. So for SEO purposes, block form pages and pages with useless junk or high html/content ratios in your robots.txt file. This will mean the pages that google does index should have better content:html ratios that will improve your overall page rank.

How to write a robots file.

+ DEBUG HTTP verb may show server debugging information. See http://msdn.microsoft.com/en-us/library/e8z01xdh%28VS.80%29.aspx for details.

I get this despite using Apache on a Linux machine with no .net/asp or other MSFT technologies on board. I can only presume its a cautionary tale told for all scans. If you know otherwise, please let me know in the comments as i could not find anything related to this on a LAMP stack. If you are running IIS though, and have .net setup, check the link in the error message for some advice on how to disable the debugging setup.

+ No CGI Directories found (use ‘-C all’ to force check all possible dirs)

If you denied from all the traffic for the cgi directories then that is the most likely message. This is a good thing. However it is best to do as it says and use the ‘-C all’ option as a precaution. This might possibly be a little brutal on your server though as it scans every possible known location for cgi directories and other stuff no doubt. It will likely take a long time, though should you find any directories still open, you best make sure you have a directive to deny from all.

The cgi directories are not the only directories you will likely need to close up. Plesk for example has a number of directories left wide open. See my other post on hardening plesk for more information on how you can fix some of the remaining issues.

Hardening LAMP / Plesk VPS / DDoS Mitigation.

If you want to harden the security of your Plesk VPS, there are a couple of things you can do from disabling non-essential apache modules to reduce attack vectors and from adding/activating some apache modules that can help in the department of security. In this article i will share with you what i have found after a couple of days of research and give you some practical tips, apache modules to install/enable/disable and some 3rd party services i have found to be of use.

What we will cover:
1) Disclaimer.
2) DNS / Proxy services (like CloudFlare).
3) Installing mod_evasive.
4) Installing mod_security.
5) Adding a rule set to mod_security courtesy of OWASP.
6) Enabling/Disabling useful Apache Modules.
7) Plesk Firewall Changes.
8) Configuring ‘httpd.conf’.
9) Disabling Directory Browsing.
10) Prevent PHP exposing itself in HTTP Headers.
11) Further Reading.

Disclaimer.

I am not a security expert by any measure, and make no guarantees that any of this will work for you! Please test your setup on a test bed before deploying and backing up your server (scripts/assets/database/configs etc) should be standard practice as i cannot be held responsible for any damages caused.

Further more, i use yum on CentOS and will not translate these instructions for use on other platforms or other package managers. If you are using apt-get, rpm, dpkg or anything else you will need to go to the original sites; the links for which i have provided and find alternative install instructions there.

CloudFlare and CDN’s.

Where DDoS attacks are concerned, there are some things you can do to help mitigate the attack but overall your options will be severely limited on any VPS. VPS’s generally mean you are already sharing your hardware with several other virtual servers, which could scale in any quantity depending on your provider and their policies.

The more virtual machines running side by side with your own VPS on the same hardware the lesser chance of successfully mitigating a DDoS attack. This is due to the limitations of bandwidth from the switch to that hardware combined with how much traffic your site and the other VPS’s on the same hardware are using.

Further more, the switch itself could have already peaked its limit, meaning nothing you do on your server will help to mitigate the attack.

If that is the case, then you have 2 options:
1) Migrate your DNS to CloudFlare or any other DNS/proxy/CDN services out there (there are lots to choose from, though CloudFlare is one of the most up and coming out there).
2) Contact your VPS provider and enquire as to any DDoS protection they are using at the switch/router level and wether they have any additional more dedicated filters that can be used. If they do provide such a service though it will likely come at a premium. Some ISPs/hosting companies will allow you to use your own hardware if you own it to filter out such attacks though will still likely charge you to use it (for the electricity and rack-space being used). Even then, such filtering boxes can be fallible.

From my own experience so far, i highly recommend trying out CloudFlare

Also, once setup on CloudFlare, considering installing their apache module from the downloads section mod_cloudflare.

Note that using CloudFlare comes with some added advantages, being a CDN, scripts/css/images and so fourth are cached, and further more, additional features can be enabled for blocking scrapers and spambots. Lastly as a really nice addition are the statistics that CloudFlare provide for you on bandwidth, and hits/visitors etc.

Now moving on to some more important tricks we can use to help mitigate weaker attacks on our system such as single DoS scripts and content scrapers etc.

Apache Modules.

Before we go ahead and do anything here, i highly recommend you back up your entire system setup, including your site contents and database and keep a copy of the backup on your local machine. I cannot be held responsible for data loss because something went wrong, so please, just backup your stuff, ok?

For this we will want 2 apache modules, mod_evasive and mod_security.

Mod_evasive will attempt to monitor connections looking for excessive repeat requests to the same resources from the same source and block out what it deems to be either DoS, DDoS, content scrapers or some other brute force attack. You can read more about that here.

Mod_security will employ a number of tactics to cover a large range of security implications from brute force attacks, xss attacks and also work to patch exploits in the OS once discovered by the community prior to the release of updates from the software providers. This is a much more sophisticated module and the original site for the project can do its description much greater justice than i can, so you can read more about it here. Though i will provide you with an extract from the site:

Just as a note here, i am using Plesk 11, HTTPD 2.2 on a CentOS setup. You may wish to tweak the install steps according to your own setup. If you do not have/use yum, you may want to use apt-get instead which is fine.

Apache Module mod_evasive.

Some of the older install instructions use the wrong download link which is no longer active, i have provided the more up to date link in the instructions for you. You will also need wget, which if you do not have on your Plesk install i suggest you install.

That should be mod_evasive installed. It is best to ensure you have not broken anything so do a restart of the HTTP Daemon.

Apache Module mod_security.

Next up, in order to install Mod_Security we need to ensure we have the following dependencies:

I am using mod_security_2.6.7 (there was a 2.7.0 prelease candidate but i don’t recommend using pre-release candidates as they may have bugs etc and 2.6.7 is considered stable).

I am also using apxs, which if you do not have/use, you may want to consult the original install instructions.

Before we can use this however, we have to edit the httpd.conf file to ensure that it is including the libxml and lua libs, use your preferred text editor to add the following at the top of the list of LoadModule section labelled (Dynamic Shared Object (DSO) Support).

If the server fails to load the httpd server, then swap the commented out LoadFile lines with the uncommented ones.

As before, it is best to ensure you have not broken anything so do a restart of the HTTP Daemon.

Adding the OWASP Rule Set.

The Open Web Application Security Project (OWASP) is a set of predefined rules to harden your mod_security setup. According to OWASP, mod_security does not do a great deal on its own, while it will by default add some protection, it will need a decent rule set to really put this thing to task. Many of the rule sets are commercial and will cost you money, however OWASP provide a nice set of rules for free to harden any server.

You can check out the project on their homepage.

Installation was a little tricky as their docs seems slightly inconsistent and poorly explained a few things. However assuming you are running Plesk 11, i will run you through the necessary steps to get this setup easily.

One of the things in the official docs was that the install directories were all wrong, or at least wrong for Apache 2.2 on a Plesk 11 setup. If this does not work for you then go ahead and follow the original docs, otherwise give this a go.

The filename extracted from the tarball may be different depending on the build/revision from the git repository.

Also, in this next section, i tried the newer 2.2.6 configuration but it seemed incompatible with the current stable build of mod_security, i could not find very good documentation explaining all the necessary steps or the difference, but my guess is that 2.2.6 is meant for the new mod_security 2.7.0 pre-release. So i went with 2.2.5 which worked. Version 2.2.6 will give you issues! (namely the httpd won’t start due to misconfigurations).

In that last step not only do we need to rename the copy to remove the “.example”, but also to change “_setup” to “_config”, this is an issue as the docs only say to remove example, and cite a wrongly named file. Just an error in their docs which i am sure will be addressed at some point but don’t let that trip you up.

Next up we need to make some changes to the file, we need to change the deny option on secDefaultAction to pass, according to the doc, (which actually will cause an error).

What they meant to say was set it the following:

To look like:

Use whatever text editor you prefer. I used vi, but feel free to use nano or pico etc.

Next up, we want to set up all the symlinks, linking the rules we wish to use into the ‘activated_rules’ directory.

I have corrected the following again from the docs to match the directories that are being used for httpd configurations in Plesk 11, which were invalid in the original docs but thats fixed in the steps i have given you below:

Now, if we peak inside the activated_rules, what we should have is a list of symlinks:

If you have something similar to that, then you are good so far.

Next up we need to edit our httpd.conf file to include the rules. This is what i added, and i added mine below the ‘extendedStatus on’ section which is only a bit below the LoadModule section. I would recommend doing the same so that it follows after the loading of mod_security.

That should load all the symlinked rule files into mod_security, and thus you should be done. :)

Now to restart the httpd server once again and check that everything is ok.

Enabling/Disabling Apache Modules.

With a complex server that is exposing itself in so many ways to the outside world, the least number of modules you can have running will help in reducing possible attack vectors that can be exploited by malicious hackers. Plesk makes enabling and disabling apache_modules super easy so we should not need to fire up the CLI/SSH any further at this point. Log into your Plesk panel and go to the “Tools & Settings” tab. There click on ‘Apache Modules’ which should be under ‘General Settings’ group.

Please check carefully the function of each apache module before turning it off, although i do not need them myself, you may depending on what your doing with your server. I have gone and made it easy by grouping them into some useful groups to explain what they are roughly about, and also provided a link for each module so you can get the official story on what its about before you go ahead and turn it off.

Here are a list of modules you should turn ON:

So, the evasive20 and the security2 modules correspond to the modules we installed earlier, if they are not on already turn them on now!

If you are running Plesk and not something else like CPanel, you will still need to be running mod_perl and mod_python even if you do not use perl or python scripts yourself on your server Plesk depends on these scripting engines (sucks i know, would love to disable them both, i just don’t use them).

Here are some modules you should turn OFF:

Filtering and substitution modules you likely will never use. (each one refiltering the same content and with potentially multiple rules being applied to each. Madness!)

Modules for remote management of database, mostly using really peculiar and obscure methods not suitable for average Plesk VPS users.

Chances are you won’t be using LDAP.

Mostly things you won’t need or care about.

With regards to deflate, please bear in mind that firstly, most browsers will cache images, css and scripts, so deflating them is pointless. Secondly as for html output, use a decent framework with template optimisation and caching for your code. For example, in php there is the Symfony framework which automatically optimises your html (strips whitespace from your html [to some limited extent, but can be optimised further with some understanding of the template engine]). Also, minify your assets manually prior to upload or during the process of building a local cache for your sites assets, this will have the same affect as using deflate except it only need be done once, where as deflate does it every single request (not good for performance!). You get extra points for using CloudFlare which acts as a CDN and will cache all your assets for you.

Next up, if you are not using SHTML SSI, CGI scripts, Tomcat/Server side Java Applets, then disable all of the following:

These are mostly tracking and logging, from a security stand point that can be a good thing, however chances are most of these logs won’t ever be checked, and all bloat your servers load time. If you need them, just pick a few but don’t enable them all. Remember that CloudFlare will block some of the damage and also log a lot of the activity for you, further more mod_evasive and mod_security will also block quite a lot of varying attacks now. I would only be activating them if i have an issue that needs a closer look, otherwise its just more crud running in the background.

You likely won’t need rpaf, if you don’t know what it is, you likely do not need it, same for mod_aclr for nginx.

Chances are you can also disable most of the authn_ and authz_ modules, unless you have specific use cases for them. The same also applies to all the proxy modules.

By disabling most/all of the above not only do we reduce the number of attack vectors but actually save a lot of memory and your website will generally load much faster. Since making these changes i have noticed nearly more than half the load time on my site. Thats nearly twice as fast. (site being CodeConsortium).

Sometimes performance is a benefit when mitigating DDoS (depending on the level of the attack), where if your site is being reduced to a crawl, at least if the content is served a bit quicker and the assets are cached on some CDN, the site may still be usable (potentially).

There may be more modules you can disable, however it is your responsibility to work that out for yourself, see what you need and don’t need. Also try them out few at a time rather than doing all of them and then finding you broke something, much easier to eliminate the potential causes from the list if you only do a few or 1 at a time.

If you want to know what modules are loaded, try ‘$ httpd -M’ to list them.

Plesk Firewall Changes.

By default, plesk leaves quite a few services exposed that you probably won’t need access to from the outside, particularly when access is already being provided through an html interface via the Plesk Panel.

Change the following to “deny incoming from all” for the following at your own discretion:

  • MySQL Server (this can be administrated through PHPMyAdmin, no need to allow direct access to this).
  • PostGreSQL Server (same as above).
  • Tomcat Administrative Interface (if your not using java on the server that is, otherwise you *might* need it).
  • Samba (Just not likely to need it, FTP will do fine).
  • Ping Service

With regard to the ping service, i choose to deny all traffic for the simple reason that, if your using CloudFlare, then your servers original IP address should be hidden behind CloudFlare. However if some malicious hacker does find your IP address, they should not be able to view the website as the traffic did not originate from cloudflare (thanks to the mod_cloudflare) so there browser should not receive any html or valid responses. So the last thing we want to do is validate their suspicions that the IP they have uncovered is legit by returning a ping request. In-fact what we want is to tell any potential hackers the least amount of useful information we can get away with; without crippling our own network.

The other thing to remember is, some of the older attacks still use ping for DDoS, denying inbound ping requests will lighten the load on the server if an attacker should choose to use this method. That said, the switch, router and even servers NIC can still be overloaded if the packet size and quantity is very large, and while we cannot really stop inbound ping based DoS attacks, at least we are not wasting outbound traffic by responding to it.

Configuring ‘httpd.conf’.

We can add some more fine grain refinements to our httpd.conf file for a little bit extra security.

1) Set, ServerSignature to Off. This will reveal less info in the http headers about your server setup.
2) Set ServerTokens to Prod. Also works to reveal less info in the http headers.
3) Add Header unset X-Powered-By.
4) Limit request body.
5) Limit XML request body.
6) Lower the timeout to 45 seconds.

I recommend only setting 1 or 2 of these at a time and then testing them before changing more. You can do this incrementally by testing each config by running ‘httpd -t’ on the CLI.

If you have already setup your CloudFlare, or any other CDN for that matter, and are using some tool to monitor your HTTP Headers. Then you may notice that most of the resources still say X-Powered-By etc while the main page html does not. This is because you need to log into your CloudFlare/CDN and purge your cache after you have saved the above settings and restarted your HTTPD server. After restarting your server and purging your CDN cache, you will notice the X-Powered-By header strings disappear.

Disabling NMAP and SendFile, is more for the instability issues that can occur. Both are designed to allegedly improve performance of serving static content, however i did not notice any decrease in performance by disabling them. This is likely because my site is entirely dynamic with the exception of assets that are already being cached on the CDN. Therefor if you use a CDN and most of your site is dynamically generated via PHP, Ruby or likewise then just turn these both off as they won’t do much for improving performance but could reduce stability issues for your OS (which is good).

Setup a sensible default for public directories that apache has access to.

Any other instances for the setup of vhosts/alias and directories etc, use -Indexes to prevent browsing of directories everywhere.

Some default apache/Plesk installs will setup an alias to /icons/, make sure that you disable this by commenting out the alias line. You can also keep the remaining setup for this by once again negating the Indexes and lastly, change the last bit to ‘Deny from all’.

Do the same again for the CGI scripts directory.

Though i added the header unset directive for the X-Powered-By, the same does not work for MS-Authored-By header string, in order to remove this, comment out the LoadModule lines for:

  • #LoadModule dav_fs_module modules/mod_dav_fs.so
  • #LoadModule dav_module modules/mod_dav.so

Some of the above tricks were found here.

Disabling Directory Browsing.

Also, you can edit your vhost.conf so that the line for Options disables indexes and allows symlinks. By doing this symlinks being active means the OS treats symlinks like regularly directories and takes less time to by passing checks on wether the folder is a symlink or not and checking if the directory it points to exists or not. This way we don’t check if it exists before trying to follow it, so it speeds things up a little. Disabling indexes will turn off directory browsing, which just adds a little extra onto our security.

Prevent PHP exposing itself in HTTP Headers.

In your php.ini turn expose_php to Off.

Remember, you will likely have quite a few versions of php.ini all over your file system, you have the default one, then sometimes another default for the CLI (which can differ), also you may have one for the Plesk system, and again another for each individual vhost.

Here is a nice script to tell you which php.ini files are exposing your server.

Output should be something like:

If you have a very short list, or none at all, then you may need to update your locate database.

Summary.

These are the steps i have taken so far, though i am continuing the search to harden the Plesk / LAMP stack. As i find new techniques that are particularly profound i will follow up with more articles. If you have any suggestions, please leave a comment below or head over to the forum. Some of the links below in the further reading section cover topics i am yet to fully investigate, and i will write more articles about other new techniques soon.

Further Reading.

Some of the resources i found which you may find useful and may include more tricks you can use not covered here.

Official / Links for Best Practices (ordered by least paranoid security approach / simplest first):
Official Apache Security Documentation
Symantec – Securing Apache: Step-by-Step.
WPSecure Server Guide.
SANS Institute – Security Consensus Operational Readiness Evaluation.
NSA Guide on Securing your OS (features documentation for most OS’s).

An auditing checklist from DISA (supposedly they do audits for the US DoD [Department Of Defence]) here (its in a zip file).

Some other links you may find interesting:
TechRepublic
20 ways to Secure your Apache Configuration.
Mad Irish: Hardening PHP from php.ini.
How To: Harden and Secure Your Linux Server – ASL and Plesk.
DDOS Protection | Sysctl protection | DDoS Deflate.
GotRoot.
Developer.Mozilla The X-Frame-Options response header MDN.

Good luck.

Custom Validators on Validation Groups in Symfony2.

If you are using validation groups and have created a custom validator, you will want to use it in your validation group, however its a bit different from how you would apply validators in the entity, and quite similar in-fact to how you define services.

Its similar to how you define services because you have 2 root nodes, first somewhat like services.yml’s parameters line, in this instance called ‘namespaces’, the second part matches the entities name which the validator will be applied to.

Starting out, your validator.yml file probably looks like this:

This is an example taken from my AttachmentBundle, which you can find here.

Now, when we have a custom validator, in this case, some validators to check a users quota on file uploads etc, we need to set a namespace for our custom validators, with the namespace parameter, and then reference that namespace below followed by our custom validator.

The name of the validator must be wrapped in double quotes so that Symfony2 is aware that this is using a custom validator, inside of which we first start the namespace, and secondly the validator alias.

We define the alias of our validator in our services.yml like so:

Authorisation on Roles in Symfony2

Following a discussion on the IRC room from someone who viewed my post about login/logout handlers in SF2, i wanted to clarify that when dealing with Roles, they should be dealt with in the controller actions.

For off, you need to make sure you got the right hierarchy of roles. In your app/config/security.yml you need something following the structure of:

Where the lowest level of access is at the top and the higher levels of access envelope the last levels of access.

Then you inside your controller actions you would do:

According the order of the role hierarchy if you have the role or higher than specified the controller action will continue, if you have a lower level of access than the minimum required then you will get an AccessDeniedException.

Git workflow.

1) create a branch.

2) add new files and commit.

3) push to github.

4) switch to new branch of github then PR and merge.

5) pull update from github with  merged changes.

Deploying SF2 Projects on Plesk VPS.

Having recently setup the new codeconsortium.com website i thought i would share some of the steps i took to get things working as it was not all smooth sailing. First off i just to clarify the platform i am using, it is a linux setup utilising Plesk and a typical LAMP setup with that.

My project written in Symfony2 and all custom code (minus FOSUserBundle), did not get off to a great start on deployment, so for other people deploying their web apps on a VPS running plesk i will share the steps i took to get things working.

First you need to obviously create your account and you can do that under Plesk control panel, this is a relatively simple step and as we are more concerned about the more complex arena of what we need to do on the command line i shall leave the creating of the account under Plesk to the documentation that it ships with.

Once you have your account setup and you have your SSH credentials handy we can begin. Firstly, download the Symfony2 framework from their site preferably WIHOUT vendors, this will greatly reduce the transfer time of things if you are going to just FTP them onto your server, but you could use git to do this also.

If you don’t have git i suggest using yum to install git. You can do this with:

Once git is installed you can download SF2 that way, otherwise just FTP SF2 onto the server, preferably inside your httpdocs directory.

The next steps we need to take is to resolve some issues with the default setup of vhost.conf, which should be located in your domains conf/vhost.conf. Once you are in your domain dir (something like /var/www/vhosts/YOUR-DOMAIN.COM/) you should find the file which you can open with the vi editor like so:

Then we need to make some additions, which i shall explain after this snippet, you need to edit your vhost.conf file to look like this (codeconsortium.com is my own domain, substitute this with your own accordingly.):

A few notes on the above snippet now, firstly, the most common issue people have is with open_basedir, which has a tendency to create permission errors with scripts, usually resulting in a blank screen with no error output. The other issue we resolve is to allow scripts and resources to be accessible through symlinks (particularly useful for assets where it is more efficient to symlink your bundles assets than copy them, particularly when you update your bundles later on). Then the section after regarding AllowOverride etc, opens up some of the remaining permissions issues. The last part of this vhost.conf allows us to use thw web folder as our actual web root directory, thusly protecting the rest of your source code (particularly your configs containing database passwords etc) from outside snooping. Once this is done, just restart apache like so:

Now that your vhost.conf configuration is out of the way, you will need to upload your project to your symfony/src/ directory on your server. You can do this either through FTP or the SCP utility on the command line, its up to you. I personally prefer using Cyberduck FTP client because you can add all manner of file types to exclude, namely large revision control files such as .git files etc. Excluding such files can cut your upload time in half in some projects with very large repositories.

Once your project is uploaded successfully, we should run the vendors install script, which you can do via:

This will install all the vendors that you did not download from the SF2 website. It makes much more sense to use the power of your VPS’s large bandwidth to download vendors than to download them to your local work machine and then use what is usually a very poor upload bandwidth of your ISP to get it all onto the server.

Before we run the check.php utility we will want to set the proper permissions on both the cache and log directories, so from your symfony/app directory do the following:

This is necessary so that symfony can write to the cache and logs directories. Once that is done, we must now run the check utility from the command line also, this helps identify any issues with your setup which will likely be a few on there. To run this type:

Following the issues highlighted modify your php.ini accordingly, usually you will do this by opening it in vi, and your default php.ini is usually in /etc/php.ini but we don’t want to edit this one, we want to edit your local php.ini which will overwrite the default one. If you choose to edit the default one found in /etc/php.ini then any changes made will effect all users on that server, depending on your root level of access on your VPS you may not actually be able to edit this anyway.

To edit your local php.ini file using vi again, edit:

OR if you do not have root access it will just be

And make all the according changes as per the issues the check.php script mentioned. Then we restart apache again:

Just to be sure that we got everything right, we can test that our settings are the right ones being used by creating a little test script in our httpdocs/symfony/web directory, create a file called phpinfo.php and edit it to look like the following:

Once all of those issues are resolved, we now want to make sure that there are no funky ‘.htaccess’ files in your root dir that might interfere with your project, so cd into your domains httpdocs directory and locate the default .htaccess directory, and delete it, if you don’t find one in there then that is good also. DO NOT remove the .htaccess file found in symfony/web you need this.

Now go to your browser and under your domain run your script, e.g; www.your-domain.com/phpinfo.php. This should output a really nice looking page with all the configs of php. Check for all the appropriate settings and check all the php.ini includes you see on the page are correct. Hopefully the chosen php.ini should have its configs overwriting the default ones, if not see what php.ini file your setup is favouring and edit that accordingly.

Now we should go back to our symfony directory and setup our entities and database stuff. Go to your symfony/app/config directory and edit your parameters.ini file so that you have input all your database credentials. If you don’t have any database credentials then you need to go into your Plesk control panel and create a new database and get your login credentials from there.

To implement this you now need to run these 2 commands (the first one you will need to run for each bundles group namespace):

That should be your database setup now, now all you have to do is install your assets. As we enabled the FollowSymLinks setting under our vhost.conf this should be no problem, just run the command below:

Now that this is done, we should be ready to test our site out, check your domain to see if you can access your site.

Hopefully all is well and you don’t have any issues, but if you do, check your error log, which you can find in your domain’s statistics directory, statistics/logs/error_log if you do not have a statistics directory then check /var/log/httpd/error_log and see what issues are coming up there. The last occuring issues will be at the very bottom of the error log, so look their first.

Assuming all is well, we can now delete the phpinfo.php script (its not good to leave this around, it can be a security hazard to leak all of your setup info)

Also remove app_dev.php and config.php from your web directory as these are only needed for dev environments.

I would also recommend once everything is working and you have tested out everything on your site that you further edit your php.ini file in your local domain area so that you set the error reporting to:

but keep ‘log_errors = On‘ as you will want a log of anything that is breaking down if someone reports it then you can trace the log to duplicate the conditions hopefully under which the bug was generated, or at least point you in the right direction.

The last step is to setup some additional tools, namely php’s APC, which is really useful for optimising your sites php performance overall. Though installing APC is not required, it can dramatically improve performance and is recommended. You can read about that in my other blog article http://www.reecefowell.com/2012/01/17/installing-apc-on-plesk/