I’m officially on mediatemple now and after getting my VPS or (dv) cleansed of plesk I noticed Yum was not installed. I’m planning on doing a full writeup after a few days but for now I’ll get this tutorial on how to install Yum on CentOS 4 out of the way.
First off you’ll need to install a key from the centos mirrors,
# rpm –import http://mirror.centos.org/centos/RPM-GPG-KEY-CentOS-4
Next let’s start installing yum,
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/yum-2.4.3-3.el4.centos.noarch.rpm
Retrieving http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/yum-2.4.3-3.el4.centos.noarch.rpm
error: Failed dependencies:
libxml2-python is needed by yum-2.4.3-3.el4.centos.noarch
python-elementtree is needed by yum-2.4.3-3.el4.centos.noarch
python-sqlite is needed by yum-2.4.3-3.el4.centos.noarch
rpm-python is needed by yum-2.4.3-3.el4.centos.noarch
urlgrabber is needed by yum-2.4.3-3.el4.centos.noarch
As you can tell there are a few dependencies that need to be installed first. If you’re going to copy and paste these commands be careful,
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/libxml2-2.6.16-10.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/libxml2-python-2.6.16-10.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/python-elementtree-1.2.6-5.el4.centos.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/sqlite-3.3.6-2.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/python-sqlite-1.1.7-1.2.1.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/rpm-python-4.3.3-22_nonptl.i386.rpm
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/python-urlgrabber-2.9.8-2.noarch.rpm
Wow that was a lot, now finally yum,
# rpm -Uvh http://mirror.centos.org/centos/4/os/i386/CentOS/RPMS/yum-2.4.3-3.el4.centos.noarch.rpm
And you should now have yum installed and ready to run. I won’t tell you how to use yum but the first thing I did was run,
# yum update
Sequential-Access drives are different than Direct-Access drives and need to be treated as such. For a while now I had always thought differently because my experience with tape was only on OS X Server. And those machines had programs that managed the tape drives for ease of use.
With a little knowledge under my belt I know differently and now I’m passing it on:
The 2.4 kernel has built-in SCSI support so you may want to upgrade. I’m using SLES 9 and 10 so I was okay.
Test to see if you’ve got it hooked up correctly and if Linux notices it:
cat /proc/scsi/scsi
Look for a non-”Direct Access” device (“Direct Access” devices are hard disks), Tape drives are primarily “Sequential-Access” devices. If it’s not there (for example, because the drive was connected, but powered on after boot) recycle the system or find a rescan-scsi-bus script to run.
cat /proc/scsi/scsi brought up this for me:
Host: scsi1 Channel: 00 Id: 05 Lun: 00
Vendor: IBM Model: ULT3580-TD2 Rev: 53Y2
Type: Sequential-Access ANSI SCSI revision: 03
To check drive status:
mt -f /dev/st1 status
although yours might be:
mt -f /dev/st0 status
Now to use the tape drive you’ll need to use programs/commands that support non-direct-access drives. The easiest which I consider ideal is one-time tar backups: The first three steps may be unnecessary, depending on drive model. Mine seems to automatically rewind and you can just let tar overwrite any existing data, rather than wait for the erase step to finish — it takes a long time to erase a 200 GB tape, trust me.
mt -f /dev/st1 rewind
mt -f /dev/st1 erase
mt -f /dev/st1 rewind
mt -f /dev/st1 lock #to prevent someone hitting eject before it's finished
tar -cvzf /dev/st1 /path/to/stuff/to/back/up/
mt -f /dev/st1 unlock
mt -f /dev/st1 eject
I already posted about using the technique of updating wordpress through SVN but it just dawned on me that the instructions I (and many others) used committed us to the 2.1 trunk. And since 2.2 is a few weeks away I’m thinking of how to transition.
If your running a few sites, the easy way would be to just run these commands:
svn switch --relocatehttp://svn.automattic.com/wordpress/branches/2.1/\http://svn.automattic.com/wordpress/branches/2.2/
Not that big transition since I can control who gets the updates and who doesn’t just in case 2.2 doesn’t play well. If I don’t use the above steps I’ll just copy my nightly script and simply modify it with the above command to run a single update for everyone in a few seconds.
The other option would be to use the wp trunk rather than the stable release repositories.
http://svn.automattic.com/wordpress/trunk/I’m not to keen with the latter but it is an option.
One of my tasks this week was to setup a web server for some sites currently hosted on the tragedy of the grid. I don’t want to talk about the (mt) grid since I’ve said so much about it before but to answer “why?” I’m still on the grid for production sites is, I’ve been too busy to migrate them and at one point I was going to put them on a spry VPS but my DB was too large for their contrants, very weird indeed and I could made it work but I pushed it aside.
This week I couldn’t wait any longer I was getting so many CS calls about it “being down” or “Sooo Slow” that it became a big priority. I was planning to install Ubuntu server and setup the system in a few hours but I had so many issues with Ubuntu Server. I rather not go into details because I can’t remember them all but one major problem was webmin and gnome. I hate to rely on a UI to manage a server but it makes my tasks so much easier. Gnome installed fine after an apt-get command but there are so many dependencies that it took hours for it to download and install. At that point I was already frustrated with the install that the problems with webmin, configuration and installing some other server packages I had givin up and started the fedora core 6 downloads.
Today I had the time to finish the system but the installation was as easy as Ubuntu except for the 5 CDs it took to install everything. That was the only 
downside for fedora over Ubuntu. But at least it’s a complete web server after install. Without phpMyadmin and webmin installed I was able to serve up whatever I needed with just a few service starts. The management tool provided as well are perfectly suited and easy to use. I can’t remember the particulars but it seemed Ubuntu wanted me to configure way to much after install without any management tools, fedora can serve a test page from apache after install and if not the tools provided make it a few clicks away.
After a few yum installs and eventually remembering to edit my host file in order to get webmin to show (from the host machine) I had a production server ready.
fedora has some great advatages over Ubuntu IMO: First SELinux (security) and server management tools. Another feature over Ubuntu right now would be the Xen support and management tools built into core 6. The Xen management will definitely help me get a couple webservers (maybe one for myself) installed with the little time I already have instead of wasting time figuring out how to install and configure it for Ubuntu.
This post is not saying fedora is a better web server then Ubuntu but for ease of use it’s clearly better out of the box.
It’s bad to remember how to flush DNS in windows and not know what to do in OS X, when it’s your main system.
Windows:
C:>ipconfig /flushdns
OS X command
bash$ lookupd -flushcache
Linux:
$/etc/rc.d/init.d/nscd restart … Read More »
February 28
I always forget these simple commands and since I just had to do a restore because of a 2.1 upgrade sna-foo.
Restoring via the command line
Restoring from FILENAME.mysql is a three step process:
1.Drop the database
mysqladmin -u USERNAME -p drop DATABASE
2. Recreate the database
mysqladmin -u USERNAME -p create DATABASE
3. Import the backup data
mysql -u USERNAME -p DATABASE < FILENAME.mysql
How to manually hardlink multiple files to point to one single file. This will drastically help you if you have multiple files that are all the same because instead of editing all of the files you just need to edit one.
This is the command that I ran
#find . -name ‘FiletoFind.txt’ -exec ln -f newfile_hardlink.txt {} \;
FiletoFind.txt is the file name that you want to search for. newFile_hardlink.txt is the file you want all the files to look at after they are found.
I then executed,
#ls -alh newFile_hardlink.txt
This will show you how many files you hardlinked to that one file.
After doing this for a work project I got, -rw-r–r– 4031 dcameron apache 1.3K Dec 15 18:45 newFile_hardlink.txt
Wow, four thousand files that I just saved myself from editing.
I’m getting my new server setup and I need to migrate all of my files over from my current one. Instead of using a rsync, which I’d recommend if in the process of migrating the old server’s files they get changed, I decided to just use tar and the scp command.
For my tar I used,
tar czvf backup.tar.gz /home/html
For the transfer you can either go from original host to new host or vice versa.
scp localfile myname@[newhost]:/[path to owned directory]
or
go to the new host and do it the other way (which I decided to do).
scp username@[oldhost]:[path to file] ./ [/path]
You could also combine the two with a |. I didn’t because I had to figure out the scp commands.
FYI: I tried to do a simple wget on the tar’d file but for some reason it gave me a 403.
