ramblings of a {sys,net}admin…

So I finally got the “ok” from the powers that be to begin the process of migrating our web servers from IIS on Windows 2000 to Apache on Red Hat Enterprise Linux. I sent out what I thought was a pretty convincing e-mail to all of those who are somewhat involved in the whole deal voicing my opinions and thoughts. I was going to include it here, but it’s reallllllly long, so I won’t.

Long story short, it worked. I was actually off work at the time I composed the message, due to a surgery I was having. After I got back, six or seven of us got together in a meeting to discuss, among other things, the pros and cons of migrating the site over to Linux. I got the okay.

Now, the last week or two I’ve been looking into various issues. I mentioned I could do this whole project on existing hardware with minimal financial costs. Now I’ve got to prove it. We had just ordered some new PCs to replace some aging ones, which proved to be good timing. I’ve now come into possession of a number of those “aging ones” and am going to be utilizing them for this project. Right now, I have two PCs that have dual 650MHz CPUs, 512MB of RAM, and plenty of drive space to host our website. Thus far this year, we’re averaging around 1.15 million hits/month, so we’re not outrageously high traffic or anything like that and these PCs should be plenty sufficient to do the job. I also have plenty of spare parts on hand in case any should die, and I can always beef up the RAM and such a bit, if necessary (I don’t think it will be).

I’m also going to be doing a bit of “rearranging” of hardware. I currently have two unused 73GB SCSI drives and a couple of servers that I’m going to “swap around”:

• a dual 866MHz server currently doing nothing, and
• a 933 MHz server running Squid and a few in-house web apps.

The plan, which involves a lot of swapping of hardware, will end up like this:

• a 933MHz server running either MySQL or PostgreSQL, and
• a dual 866MHz server running Squid.

The server running Squid is by far the one that’s the “busiest”, as we have users (students, mostly) surfing the web all day long (duh!). It has the highest load average of any of them and will benefit the most (I think) from the extra CPU power. Then, the 933MHz server will be a dedicated database server to support the databases used by our website (which is currently completely static — more on that later) and other databases used internally (most of our database stuff is on Microsoft SQL Server).

There are a few things that I was still kinda unsure about, but that was pretty much taken care of today. The two PCs that will be running the web site will be using heartbeat in an active/passive failover mode. That way, if the primary server goes down or otherwise becomes unavailable, the other will step up and take its spot. This should help me sleep a bit better as I don’t have to worry so much about when one of them dies. Now, if they both at the same time, then we have issues. =)

Today I got the two PCs set up identically with most things set up before I “imaged” the second one (Apache, postfix, etc.) so those are ready to go. I just installed heartbeat tonight and I’m actually thinking about going back in tonight to install an extra NIC in each one (for monitoring with heartbeat — crossover cable, baby!) and hookup a modem cable between ‘em. Of course, I’ll have each of their NICs plugged into separate switches and such to help provide a bit more redundancy.

So, I still have some work ahead of me, but I have faith that everything is going to work out well. Oh, I mentioned earlier that our web site is completely static. If anything on it changes, it’s because someone changed it by hand. This wouldn’t be so bad if we didn’t have “dynamic content” on our site. Take our little “events calendar”, for example. Anytime a new event is announced or coming up or what have you, $webmaster has to go in and manually update the site. “Wouldn’t it be nice if he didn’t have to?” was one of the questions I asked in my initial persuasive e-mail. The thought being that we can set up a front-end web app so that authorized users can add their own events to the database backend. Then, whenever the web page is accessed, it will automatically pull the latest event info from the database. Common sense, yes, but evidentally it’s not so common. Oh, and have I mentioned that I hate ASP? Well, I do, and I won’t write any if I can avoid it. At least if it’s on Linux, I have my choice of C, Perl, PHP, etc. I shudder when I think about having to do this in ASP. Oh, and the same thing for podcasts. We just “launched” our first official podcast, and will have more coming up. It’d be nice if the right people could access a front-end to upload their MP3 files, provide the info (title, description, etc.) of the podcast and have it automatically made available via our XML feed. Once again, common sense.

So, long story short (have I said that already?), this is going to be a pretty fun project to get going. I’m a huge FOSS freak, so I view it as a victory for the open-source world. =) Stay tuned, and I’ll keep you updated on how it’s going.