Last week a friend of mine with a slow dial-up internet connection asked how he could download all of the newest updates for his FC1 system and take them home on a CD. I showed him how.
A small cluster of a router, one server and two workstations all have FC1 installed. Why not save upstream bandwidth by locally mirroring all updates? That's just what I did.
Let's get started. What you'll need: a server with httpd and plenty of extra hard-drive space (At least 5 GB). To retrieve the packages you'll need the rsync client. To automate the process you'll need the cron daemon. And of course, you'll need to have yum on your clients.
I happen to have a directory on my server called /space. It's a user-writable location that doesn't get included in my weekly backups like /home does. I'll create a directory /space/mirror for all of my mirrors. Then I softlink the directory so that it shows up on my webserver.
$ mkdir -p /space/mirror # ln -s /space/mirror /var/www/html/ $ ls -l /var/www/html/mirror lrwxrwxrwx 1 root root 18 Jan 25 03:04 /var/www/html/mirror -> \ /space/mirror
(I'm doing all of the following steps as a non-root user. Safer that way.)
Now I'll write a script that does the syncing. The script doesn't do much but call into rsync to do all of the heavy lifting. I wrote mirrors.kernel.org here, but you will want to find a closer mirror.
#!/bin/sh # mirror-script.sh # hal canary DATE=`/bin/date +%Y-%m-%d` OUTDIR='/tmp' MIRROR=/space/mirror [ -d $OUTDIR ] || mkdir -p $OUTDIR OUTFILE=$OUTDIR/mirror-output-$DATE.txt /bin/nice /usr/bin/rsync --verbose --progress \ --stats --archive --partial \ --exclude development/ \ --exclude test/ \ --exclude 1/SRPMS/ \ --exclude 1/i386/iso/yarrow-SRPMS-disc1.iso \ --exclude 1/i386/iso/yarrow-SRPMS-disc2.iso \ --exclude 1/i386/iso/yarrow-SRPMS-disc3.iso \ --exclude 1/i386/debug/ \ --exclude updates/testing/ \ --exclude updates/1/SRPMS \ --exclude updates/1/i386/SRPMS/ \ --exclude updates/1/i386/debug/ \ mirrors.kernel.org::fedora/core/ $MIRROR/fedora/core/ \ >> $OUTFILE
You'll notice that most of the arguments are exclude statements. I don't need to sync any of the source rpms, so I'll leave those out.
Now make that file executable and execute it.
$ chmod +x mirror-script.sh $ ./mirror-script.sh
It will take some time to download all of those files. Come back in an hour or two. you can use 'tail -f mirror-output-<date>.txt' to watch the progress.
Once you're done, try browsing to http://[YOUR_SERVER_NAME]/mirror to check that it works. (Assuming your http daemon is up and running.)
I made a file called $HOME/.crontab that contains the line:
15 1 * * * /space/mirror/mirror-script.sh
You can edit this file all you want. This will run the script at 1:15am every day. Then I ran the crontab program to make this a cron job:
$ crontab ~/.crontab
'crontab -l' tells you what jobs the cron daemon will do for a particular user.
$ crontab -l # DO NOT EDIT THIS FILE - edit the master and reinstall. # (/home/hal/.crontab installed on Sun Jan 25 15:24:10 2004) # (Cron version -- $Id: crontab.c,v 2.13 1994/01/17 03:20:37 ... 15 1 * * * /space/mirror/mirror-script.sh
This should burn all contents of the update directory onto a CD, assuming cdrecord is setup (look at /etc/cdrecord.conf).
$ mkisofs -r -J \ -V "Fedora Updates as of `/bin/date +%Y-%m-%d`" \ /space/mirror/fedora/core/updates/1/i386/ \ | cdrecord -v -
That should be it.
Now, (as root) edit your /etc/yum.conf file on each machine you want to look at the mirror
[base] name=Fedora Core $releasever - $basearch - Base baseurl=http://[YOUR_SERVER_NAME]/mirror/fedora/core/1/i386/os/ gpgcheck=1 [updates-released] name=Fedora Core $releasever - $basearch - Released Updates baseurl=http://[YOUR_SERVER_NAME]/mirror/fedora/core/updates/1/i386/ gpgcheck=1
Then do an update just to check it out:
# yum update
FN Site Search
FN Blog Login