Installed Ubuntu

I tracked down a spare 8.5gig disk today (the one that came with my old P2-300 box, ironically) and installed Ubuntu on it. First problem: I installed the spare disk as an IDE slave, and the Ubuntu install totally hosed the boot loader on the master (Grub). After installation, the boot loader gave some cryptic error message and hung. So, I booted into Knoppix and reinstalled the boot loader, which allowed me to boot into my Debian OS on the master disk. I then attempted to configure my existing Grub to boot Ubuntu off the slave. But, when I booted, grub refused to recognize the slave disk. Not sure why (BIOS issue maybe?) but I ended up copying all of the Ubuntu /boot stuff into the /boot partition on my master disk, pointing grub at that, and just booting everything from there. Once I did that I was finally able to boot Ubuntu. (One hitch with this method — Kernel upgrades in Ubuntu are no longer automatic. I have to copy the kernel and initrd images into /boot on the main disk, then edit the gruf.conf there to boot the new kernel. Not a big deal, as I don’t plan on running this configuration for too long — if I like Ubuntu, I’ll install it as the main OS on the computer.)

Upon bootup, it immediately prompted me to install 160-odd megs of software updates, which mostly worked, but some of them apparently crapped out as I got this happy-fun-ball “some updates failed to install” window after the installation finished. Being that Ubuntu uses apt, this is somewhat to be expected, but I hope it doesn’t screw up further updates (as apt failures on my Debian boxes are wont to do). Followup — no further problems with apt so far. After installing the updates, I was prompted to reboot, which I did, which brings me to where I am now, writing this entry.

Ubuntu seems nice enough, but so far it doesn’t seem much different from other Linux desktop installations I’ve seen, all of which are fraught with quality-control issues such as these. Once configured, they work well, but there’s always that pain of setup and configuration. I guess I’m a little disappointed — after all the hype I’ve read, I was hoping Ubuntu would be more revolutionary — a Linux desktop that doesn’t really feel like a Linux desktop. Oh well. Off I go to a command-line to get my graphics card working and fix my fonts, just like every other Linux desktop….

[More:]

OK.. Installing the nvidia driver was easy, actually. There’s a package (nvidia-glx) that takes care of it. After installing this, I went in and copied my configuration out of my old xorg.conf, restarted X (by way of the gdm display manager), and it came right up with my dualhead configuration.

I’m now in the process of installing some other “must-have” apps such as emacs, thunderbird, etc. Oh yeah.. and OpenAFS. Uh-oh…

Well, openafs turned out to be painless. Just install the modules-source package and follow the directions in /usr/share/doc/openafs-client/README.modules. Now to work on fonts. Installing msttcorefonts package helped a lot. To do that I first needed to go into Synaptic (Ubuntu’s GUI front-end to apt) and enable the “multiverse” repository, which includes non-free packages. Then, I found that my X display was not set to 96x96dpi, which is supposedly optimal for font rendering. Based on info found here and here, I tweaked the display DPI in xorg.conf by adding the following option to my “Monitor” section:

DisplaySize 783 277 # 96 DPI @ 2960x1050

and the following to “Device”:

Option "UseEdidDpi" "false"

Next it looks like I need to tweak anti-aliasing for certain fonts (reference).

Little by little it’s coming along.

Another good font tutorial for configuring fonts under Ubuntu. This one includes instructions for installing the Tahoma family, which for some reason is not included with Microsoft’s Core fonts for the web. With the MS fonts (plus Tahoma) installed, things look much better already, and apparently I can improve the look further by tweaking anti-aliasing and other stuff… might play with that a bit tomorrow.

First impressions of Ubuntu, etc.

Last Friday I tried out the latest release of Ubuntu Linux. They provide a “live” CD, which boots and runs directly from the CD just like Knoppix. My goal is to find a nice desktop-oriented version of Linux that “just works”. On the server side, I’m sticking with Debian, but I find vanilla Debian a bit lacking in the desktop department. So as a stop-gap between cutting over to OS X completely, I thought I’d try out Ubuntu and see how I like it. Ubuntu is based on the same apt package system as Debian, so it’s familiar, and it’s touted as being very desktop-friendly.

First impressions: it looks nice. apt-get works as expected from the command line, but the default archive server has a very slow connection — I wonder if there are mirrors on Internet2 that I could use. If not, that’s a definite drawback, as I’m not sure I could give up the blazing speed I get from debian.lcs.mit.edu. I was able to install xmms easily, and my sound card was immediately recognized, and the system shares the sound card between apps. However, for some reason it didn’t work streaming MP3s from my mp3act server. Recent versions of OpenOffice and Firefox are provided. It didn’t pick up my dual-head setup, but I didn’t expect it to — I’ll need to download and install nVidia’s x.org driver manually. It looks like I’ll need to install some compilers and devel tools before I’ll be able to build the nVidia driver. But I expect it’ll work.

As with every other version of Linux, the default fonts are butt-ugly. Why can’t someone put out a Linux distro that has nice fonts out of the box? That has always been one of my biggest gripes with Linux. There are tutorials on the ‘net to improve the look of the fonts under Ubuntu, but honestly, this shouldn’t be something I have to mess with. Linux is never going to get anywhere in the desktop market until they can get past this issue.

All of that said, I think I may try out an “official” install of Ubuntu on the hard drive, and see how it goes for awhile. I’d rather not wipe out my existing Debian install, so I’ll have to scrounge around for a spare hard drive first.

In other news.. I’m thinking about finally taking the plunge and going with totally paperless bills and financial statements (where possible). My redundant disk setup gives me a more reliable place to store documents electronically, so there’s no reason not to go for it. As with everything else, I’ll see how it goes.

MySQL Replication

I’m about done with my computer shuffling which I started a month or so ago. I have a 300g drive at work and a 120g drive at home. The idea is to replicate stuff like the MP3 collection, photos, system backups, etc. in both places, to guard against losing data to a disk crash.

The next challenge is to set up an mp3act server at home that replicates the one at work. I’m going to try doing this with MySQL replication. The idea here is to replicate the mostly-static music library tables with work being the master and home being the slave. Then, each site would have its own copies of the dynamic tables like playlists and playlist history. This lets me use one-way replication and avoid setting up a dual master/slave configuration, which I don’t think would work well with the dynamic tables, particularly WRT simultaneous access etc.

Yesterday and this morning I took the first steps toward doing this, which involved locking down my MySQL server at home, then setting it up to allow remote TCP connections (comment out the bind_address option in my.cnf). Then I needed to create a slave user, which I did by doing

GRANT REPLICATION SLAVE ON *.* TO slave@homedsl IDENTIFIED BY 'password'

The grant command will create the user if it doesn’t already exist.

Then, I needed to edit /etc/mysql/my.cnf on both the master and the slave, to give each one a unique server ID. I gave the master an ID of 1 and the slave an ID of 2. This is accomplished in the [mysql] section of the config file, with a line like this:

server-id=1

Next, I followed the instructions in the documentation (see link above) to lock the tables on the master and create tar archives of each of the databases I wanted to replicate (on my Debian box, each database has its own directory under /var/lib/mysql). I then untarred the databases into /var/lib/mysql on the slave. For each database I was copying, I added a line like this to the slave’s config file:

replicate-do-db=database-name

I found that if I wasn’t replicating all the databases from the master, the replication would fail unless I explicitly listed the databases like this. The slave expects the tables it’s replicating to already be present — it does not “automagically” create tables as part of the replication process. I wasn’t really clear on this point after reading the documentation; it was only clear after I actually tried it.

With the tables copied over and the configurations tweaked accordingly, I followed the docs to record the master’s ‘state’ info, point the slave at the master, and start the replication threads on the slave. This all worked as expected. All in all, it wasn’t too hard to set up, so I’ll see how well it works going forward.

Whew… what a day

Thursday was quite the day of triumphs and tribulations.

It all started out with a successful swap-out of my home Linux server. It had been running on an old, trusty-but-tired Dell P2-300mhz, and I upgraded it to a slightly-less-old Dell P3-450mhz (Linux is far from perfect, but it truly shines at getting new life out of old, scavenged hardware). The upgrade was as easy as: build a new kernel, shut the old box down, pull out all the boards and peripherals, put the stuff in the new box, and boot the new box up. The result was a home server with a 50% faster processor and an extra 256mb RAM (640mb total vs 384mb). Not earth shattering, but a noticeable improvement, and it was easy to do. The trick to doing is this to transplant as much of the “guts” from the old box to the new box as possible, so the hardware configuration stays mostly the same.

Next up was the launch of our new myUMBC portal, which so far has been very smooth, other than the usual little niggling problems that always pop up with these things. We had already been running uPortal in production for six months, and that experience definitely helped ease the transition. The centerpiece of this launch was a complete redesign of the UI, but behind the scenes we also upgraded the framework and layout manager. This gave us an opportunity to start out “clean” with a fresh set of database tables, and to redo a few things which were done sloppily with the initial launch (such as our PAGS group and channel hierarchies). It positions us very well for future releases and gives us a clean platform on which to build future improvements.

Of course, by Murphy’s Law, the air conditioning in ECS picked yesterday to go on the fritz. So the launch was a little sweaty, but it happened anyhow. When I left the office, the A/C was finally getting back to normal, and then I get home and our power goes out. It ended up being out for around 4 hours, from 7:30pm till 11:30pm. Not all that long, but it always seems like forever when it’s actually out, and it made for a pretty sweaty (and surly) day. And of course, BGE’s automated system never gave us an estimate of when the power would be restored, so Cathy packed up the kids to go sleep at her sister’s, and pulled out 2 minutes before the power came back on. Fortunately I was eventually able to get a decent night’s sleep. I must say I’m more than ready for this summer to end. This is the first truly miserable summer we’ve had since 2002, and I had forgotten just how bad 2002 was. Oh well… t-minus 7 weeks till the Outer Banks.

CWebProxy channels and self-signed certs

OK. Just so I don’t forget this when I inevitably have to do it again.

We are starting to add some CWebProxy channels that access the portal web server via its external URL rather than one of the loopback interfaces (long story why, but there are a few issues with proxying to a localhost URL, particuraly WRT inline images). These channels go through SSL, as opposed to the loopback ones which use standard HTTP. Our test portal server uses a self-signed SSL cert. That causes some problems, because the portal doesn’t have access to the server’s cert to properly negotiate the SSL connection.

Solution: Create a local keystore containing the cert info, and point the JVM at this file via a command-line argument.

How to do it in 5 easy steps:

  1. Find the SSL cert for the web server. On the portal servers, this is located under server-root/conf/server-name.crt. Make a temporary copy of this file. Edit the copy and remove all lines except the actual cert data, including the -----BEGIN CERTIFICATE----- and -----END CERTIFICATE----- lines.
  2. Use the cert file to create a Java keystore file. Assuming the keystore will live at /etc/umbc/uportal-test.umbc.edu.keystore and the cert file copy is cert.txt:

    keytool -import -trustcacerts -keystore /etc/umbc/uportal-test.umbc.edu.keystore -file cert.txt -alias uportal-test

    (Note: keytool is in JAVA_HOME/bin on recent versions of the Sun JVM.)

  3. Set permissions on the keystore file so that the portal web server can read it.
  4. Point the portal web server’s JVM at the custom keystore file. With Tomcat, this is done by setting the JAVA_OPTS environment variable prior to starting Tomcat. For UMBC web servers, the place to set this is server-root/bin/config-perl.
  5. Restart Tomcat.