I have been working with open sourced software like this for years so I expected some bumps but figured in a few hours of free time I should have it up and running. I was so utterly wrong I could not have imagined the pain this software would cause. First what little documentation there is pre supposes configuration options and package installations that are almost sure to be incorrect for anyone who attempts the install. Second the server makes use of some new technologies like extended attributes for files that while mentioned are not described in enough detail for anyone not already intimately familiar with to make use of as for most installations this is disabled by default. So the setup isn’t the easiest to grind through. The good news is if you know what hurdles await there really is nothing to it.

So here are some of the steps required to get it up and running under Ubuntu 7.10.

My assumptions are that you are at least running a base installation of ubuntu and that you have not removed any of the packages installed by default and are using the default ext3 file system. I will also assume you understand how to deal with any issues regarding apt-get and mirrors.

First you want to get any extras that might not be installed by default with the following command.

apt-get install curl zope3 python-xml python-pyopenssl python-dateutil python-xattr python-pysqlite2 python-twisted python-vobject python-kerberos

Next you want to setup the file system for the metadata storage. This part took me quite a while to figure out that it was indeed a showstopper not to have a file system that supports this interaction. Fortunately I was running ext3 so it was a simple as running the following command mount -o remount,user_xattr /. This remounts the root partition with the ability to read and write the required metadata. This is required for whatever partition you store the data on only and once you have done this you may want to add user_xattr to your /etc/fstab in the options section.

Once this has completed you are ready to get the server from svn. You can place this any number of places popular ones include /opt /var and /usr/local so change your directory accordingly and run the following command.

svn checkout http://svn.calendarserver.org/repository/calendarserver/CalendarServer/trunk CalendarServer

Keep in mind not everyone is a trunk kind of person you can also use the release versions found by pointing your browser to http://svn.calendarserver.org/repository/calendarserver/CalendarServer/tags/release/ and grabbing one of the directories you see in there and use that url in place of the one pointing to trunk.

After the files have finished downloading it is time to configure your server defaults you can do this by changing directory into the Calendar Server directory and running the command run -s once this has completed you need to create your config file so cd into conf and the run cp caldavd-test.plist caldavd-dev.plist. I don’t know why that was their naming convention but it is.

Now you have a working server ready to accept local connections to enable remote connections edit the caldavd-dev.plist and search for 127.0.0.1 and replace it with the external interface you wish the service to listen on. You can now configure your client to access the server to test out prior to getting into the config file and securing it for your personal needs.

The client connection from iCal was pretty simple you go to the iCal menu select preferences. Then you select the accounts tab click the plus sign to add a new account. Type in your own description and then the username and password to connect to the server. Next enter the url to the server begining with http or https depending on your setup. Then the hostname or ip of the server you are connecting to. Leave off the trailing / it will auto setup your path if you elave it off. Click add and you should be reading your online calendar.

Some things I hope to see in the future are event notifications when something has changed on the server it should notify the client to check for modifications. It should also maintain the connection to the server when iCal is closed and update notifications accordingly. Overall thought I’m happy with how it does work and glad I took the time to get it setup.

I learned quite a bit in this little test. The most important thing you can never be to prepared. When I started I thought I had everything laid out and I had accounted for all the potential problems. What I ended up with are problems I had never encountered before and could not have predicted. Should there ever be a next time I will simply budget in the time to deal with the unexpected instead of trying to plan for it. It really did nothing more than given me a false sense of security.

I also learned that when moving from one distro to another for many of the configurations. Generally your better off reconfiguring everything from scratch vs. trying to port old configuration to the new system. the only config files that made the migration consisted of key value pairs that feed various features in postfix like the aliases db. Attempting to cut that corner ended up costing me a lot more time and frustration.

Spam assassin under Ubuntu is pretty easy if you know to do the following to the config file.

After the smtp line at the beginning of the file add the following on a new line.

-o content_filter=spamassassin

Then at the very bottom of the file define the filter with the following

spamassassin unix – n n – – pipe
user=filter argv=/usr/bin/spamc -f -e
/usr/sbin/sendmail -oi -f ${sender} ${recipient}

This will get spamassassin processing as expected then you can work with the config files in spamassassin. I’m not sure why it doesn’t do this by default but it didn’t so thats the fix.

Another issue I had with postfix was

warning: connect to transport filter: Connection refused

This is what prevented me from being able to connect. Getting postfix up and running is difficult unless your very methodical with your config. The easiest way to trouble shoot something like this, as I noticed many others with similar issues while trying to figure it out for myself, is to disable configuration options one at a time. Consistently disabling half the remaining config options might be faster but one at a time will help you narrow it and be more certain about it.

My solution ended up being a missing file in the postfix directory for mynetworks. This is what was causing me to have to manually deliver each message only took 7 hours of dedicated reconfiguration to discover this little tidbit.

All in all I’m glad to be done with this I’m excited to play with some of the new technologies this setup will allow me to play with instead of the lost time configuring the day to day stuff it needs to handle.

Oh and if your looking for virtualization solutions VirtualBox seem rock solid, and VM Ware is great. Parallels kind of did a bad thing to their customers this month and while they make a good product its not worth the potential to get burned by a company that has shown they don’t really care about their customers, just their customers money.

Here is the interface map

br0 216.87.144.154
br0:1 216.87.144.155
br0:2 216.87.144.156
br0:3 216.87.144.157
br0:4 216.87.144.158

br0:2 and br0:4 are unable to communicate with the internet in any way shape or form. The rest br0 br0:1 br0:3 all send and receive traffic just fine. But it isn’t linked to the interface, after further testing I found that the ips ending in .156 and .158 are unable to communicate with the network. These used to be the interface ips I used to keep mail and web separate and where the highest traffic ips on the box so I figured something with my provider kept these ips from working so I called them up for resolution. While on hold I decided to fire up the windows VM and assign it one of the broken ips. A quick ping and it was alive. I tried the same with the 158 ip and it worked as well. So now what I have is something I can only imagine is a bug in the bridging code that is preventing those two ips from working on a virtual interface. I’m hopping someone who reads this blog can throw me a bone. I have spent all day with it and am at an utter loss.

For the records iptables where disabled, all proc options where returned to defaults, the route was checked for an inadvertent bitmask, all virtual interfaces where removed except for one and it was assigned to one of the dead ips and it wouldn’t work changing it to any of the ips that had worked previously worked fine. Multiple reboots later and here we are, windows gets 156, and 158 the rest I’m leaving with the host os until this is all sorted out.

Aside from the previous issue everything went smooth, I was forced to abandon the original setup in favor of a single linux host, with a windows server guest. So far the configuration is most stable like this. Using some fancy shell scripting, I will post later, I have a VM that starts and stops with the system and is able to save the machine state between reboots. This is great for those moments when I forget about the VM and simply issue a reboot command to the host.

Here is my pretty picture to help me keep things straight. So far everything seems to work fine and I even was able to pull up a backup of this site without issue. My main concern speed is a non issue, moving from a 1Ghz with 256MB of ram to a dual core 3Ghz system with 2GB of ram is plenty to run all three oses faster than I was running the one machine before.

Later I will use the handy yum remove command to handle all the deps when removing all those packages but I thought they might be handy during the install. I”m going to try and get as much done tonight as things have come up that might preclude me from finishing this as quickly as I would like to.

I’m also looking for cheap win 2k3 os that will allow installation of SQL server 2k5 developer edition. If anyone has any sources of that hit me up via the comments. Right now I will continue to use my win xp pro install as it’s the only license I have that has iis but I would like to be able to handle multiple virtual domains in the future without any weird restrictions on what I can have installed on the box.

Here are the results of the two VMs starting up in their new home. Something tells me I missed something somewhere.

First step is to get freetds, and freetds-devel installed from livna.

rpm -Uhv http://rpm.livna.org/livna-release-6.rpm

this will download and install the repository. Once installed

yum install freetds-devel php-devel

This will get the packages installed that will be needed to build the php-mssql module if you have all the rest of the appropriate developer utilities installed. Next get the php-nonfree src rpm file from the remi repo at

http://remi.collet.free.fr/rpms/extras/

Also, don’t worry about the minor version number just make sure if you are running fc6 to get the fc6 version. Once you have the src rpm file downloaded run

rpmbuild –rebuild php-nonfree-5.1.6-1.fc6.src.rpm

Once this finishes if there are any dep errors you will have to resolve those as appropriate. If everything goes well and the process completes you need to check the /usr/redhat/RPMS/i386 folder for the php-mssql module. Once you find it run

rpm -Uhv php-mssql-5.1.6-1.i386.rpm

It should auto restart apache and you should be in business. Modify the /etc/freetds.conf file to include your server so you can reference it by name. I can’t for the life of me figure out why this module isn’t just included. I guess so long as it works like this I’m Ok with it but it would be nice for an officially supported release to be available.

This means at least this part of the project in moving to the new server will be a little less painful than anticipated. Here is for everything else going smooth from here out.

Crazy and just a small part of what the technology will offer. I have a server on the way I plan to use this same technique on to run Linux, and Windows servers on for my personal use and development. While playing the hurry up and wait game that I usually play when waiting for hardware, I realized if I’m going to virtualize these systems later anyway why not go ahead and find a good VM that runs under linux and get to setting up the server.

So I did I found Quemu, KVM, Xen, Parallels, VMWare, and the rest of the usual suspects. Then a weird name caught my eye something i had not read a million times before about. Innotek (Office Space Reference?) VirtualBox. They had a mac version so I downloaded it and tried it out. I used an ubuntu iso that gave me some trouble under Parallels and it work flawlessly. It even seems to run a bit faster than Parallels.

I spent quite a bit of time testing this setup to make sure it would work for my use and then I downloaded the manual. To my surprise the VM supports RDC. This means without any other software installed in the guest os I have remote control of it. In addition there is a headless mode that is perfect for servers that aren’t running a gui.

Image of Fedora Core 6 Install

This and the lack of any additional monetary investments made this product the perfect test candidate for my little experiment it will be interesting to see how the next few weeks play out while setting up this new server. Luckily I already have the linux side finished and I will finish up the windows side tonight. This means when deployment time comes I need to get a base linux install, put the VirtualBox software on the host machine copy the guest oses to their new home and create a startup script for start the guest oses when the machine boots and I’m ready to deploy.

Then I will have two systems running side by side and complete control over how they interact and if I want to later I can change them out in the field entirely remotely.

The new box will be running a mixture of Windows Server 2003 and Fedora Core 7 which should be out in the next 2 weeks. Until then this is just a pace holder and probably won’t be updated until I finish the server setup. Once it is done I imagine about 1 week to get all my dev tools installed and running on the server. By June 1st the new site will be up and I should have all my old blog posts ported to the new server.

First thing I noticed was the fact it rebooted my phone over and over and over again until finally I removed the update images from my webserver.  Kind of anoying but par for the course for Grandstream.  It wasn’t until after i had gotten past this I noticed this firmware isn’t your normal Grandstream POS firmware written by two guys who still watch beavis and butthead episodes while trying to fart in each others faces.  This actually looks like some thought was put into it.

Ladies and gentlemen we have DND from the phone itslef without any crazy scripts or wierd dialplan entries we can control DND from the phone itslef now.  The indicator is a little retarded at the top of the screen but it is much better than what we had before.

First now when the bootup happens you can see on the phone itself when it is provisioning, downloading rom files, writting rom files, getting the ringers etc…  I like this part becuase it makes it easy to see when the phone is locked up although I have not seen that yet with the new firmware.   Then once the phone is up I did nottice some apparent improvements in audio quality.  They weren’t mind blowing but the person on the other end of my test call agreed.  Next when I spend a few more minutes with the phone I noticed the menu looks like a real menu system instead of some hack slapped together.  Real scroll bars etc…   Some of the more usefull features are a phone book and real call log with a history of recieved calls and missed calls.  I noticed that if you miss a call the display will show y9ou that you have x number of missed calls.  This is something that has been sorely missing.  In adition there are many more controls in the web interface including controls for using the phone as a NAT router for DSL?  Its in there why you would use your VOIP phone like that instead of just buying a cheapo dedicated box I don’t know but hey its an added feature with very simple and primitive controls ATM.

Inside the web conf most things are still there I couldn’t find anything that I esed that was missing anyway.  Some additions are for managing Updates whic are not very clear how they work.  Under basic setting we have a new option for speed dial keys BLF.  This means wen can setup those little red LEDs to monitor pretty much whatever on asterisk and turn them on via notification events that have been subscribed to by the phone.  This can make for some interesting setups where you could actually see how many lines are in use on the phone itself or possibly monitor other events with some creative dial plans.

Other than the above mentioned the only other noteworthy addition to the webconfig was the turn off speaker on remote disconnect.  This means no more anoying busy signal when the remote party disconnects.  The speaker phone just hangs up like it should.

So far I have only come across one anoying thing.  The routines for handling the screen redraw seem to be buggy leaving artifacts everywhere at certain times.  But my over all opinion is this may be the firmware that redeams Grandstream in my eyes at least.  To me this is the firmware they should have started out with as it really.

If you have a GXP-2000 and are looking for a way to make it work like a real enterprise phone should the next version of this firmware very well could do the job and the current one shows us they are finally on the right track.

I’m going to attempt to upload some video showing some of the new interface and I will appologize in advance for the poor quality.