Archive for August, 2005

Something else today

Monday, August 15th, 2005

Well, during the weekend, I got rather lazy. I haven’t done half the things I’ve had in mind, so I tried to spend some of today for the sake of these things. Nada. Wasted my time on work, and work, and then some more work, I’ve done nothing useful whatsoever.

During night time, I’ve decided I’d do these things yet. So far, one of my major
capital projects would start by me installing Linux on a computer lend to me by a friend. Easy, you say. True indeed. However, I’m out of blank CDs, I’ve got FC3 extracted files on one of my network shares (nfs, of course), and I have a bootable cd of FC2. No go. I’ve tried making the poor on-board sis network card netboot, and already have full bootnet (pxe) supporting environment, but nada. Nothing. Tailing the logs of the DHCP server, the client doesn’t even bother querying for IP address. It just won’t work. Adding another Davicom NIC (DM9102AF), proved to be a waste of time as well. The card wasn’t fully operational. I will just have to buy new blank CDs…

Searching somewhat in Google, I might be able to solve it, just with an additional directive, or so. If so, I will publish here my relevant configs.

Not much of a programmer

Thursday, August 11th, 2005

I’ve never been much of a coder. I used to write a little pieces of code in C, when I was student, but long since I’ve stopped.

I have had a resolution just now. To learn PHP and Perl. I want to build few things, and now is as good a time as ever.

I have the literature, I have the will, I only need enough of it, and some time.

Linux Vservers, and stuff

Wednesday, August 10th, 2005

I’ve had the "pleasure" of reconfiguring Linux Vservers (virtual servers, kernel level, etc), to add additional IP addresses to each. Although this setup is old, and I wouldn’t dare change it to anything else (not on a production server, running 3 vservers, each running 400 hosted sites, etc), so I’ve checked the scripts, and so found out few bugs. For example, someone thought that running X=`expr $X + 1` would do the trick. Much easier (and working) is to do let X+=1. Right?

So, I’ve discovered a bug there, which prevented the usage of more than a single virtual NIC for each vserver, and fixed it. I’ve also started fixing the remains of a lousy setup, where everything is explicitly described, aka, you have conf files, which include the directive "ONBOOT=yes", and you have to put in the startup script the names of the files to read. I rearranged it to scan the directory holding the conf files, and based on the ONBOOT directive, to either start the server, or not.

Also, each Vserver was limited to 16 virtual IP addresses. Maybe it’s with old vservers, and maybe it’s a total limitation of the system, unsolvable, but the customer won’t get his 128 IPs per vserver…

I hate lousy script jobs. I hate
over-customization of servers, where you’ve defined one server, and you have to manually start defining each one of the next 30. I wish I could kill the person who’s done that. It sux.

System hang, and the results

Tuesday, August 9th, 2005

Well, the results weren’t for the hang, but were for the long while since it was last rebooted. Usually it’s around 2+ month between reboots, my desktop, and it leaves you with tons of questions as to what might happen when you actually boot up. Well, it never happens when you want it, or have the time for it…

So? What did we have? A system which hung. probably because I’ve used Disk-On-Key with it. On some occasions I’ve seen it happening. Weird, because last time I’ve seen it, it happened on my laptop, running older kernel, with swsusp running. I don’t use sesusp on my desktop…

So the system allowed itself to boot up. And waited long enough (around 1/2 a minute!) while bringing up networking for sendmail (why? I’ll have to look into it), and then GDM decided to crush. It just kept on crushing. Debug mode did not reveal anything I could work with. After a while (which included testing of X, and re-running gdm, etc), I’ve decided to use any other DM which can
work. I’ve used XDM, and it took me only a short while to make sure it allowed my favorite desktop. Wasn’t nice, though.

I think they have some issue with GDM for Debian systems. I don’t know where it broke, but it used to happen on my laptop as well. Now it has stopped doing it, for some unknown reason, which means I will have to compare package versions.

On another note, and a post I’ve tried to add here, just before my computer froze. I’ve had a BlueTooth headset – Motorola HS805 for a short while now. This BT device uses AAA batteries, which means it’s a bit large, but can hold lots of time. Moreover, it’s design is one which will not penetrate one’s ear canal, but will remain outside of it, aka, external to the ear itself. It’s good. You don’t feel like someone’s poking his finger into your ear. It also has a good sound quality, which makes it a good product. However, one flaw which I couldn’t ignore rendered two such devices as useless (and broken) for me. The battery compartment cover is being held shut by two small plastic pins, or holders. Tiny as they are, holding the whole might of the battery and the spring at which it is pressed, leaves, at least for me, a very high chance of breaking one of those pins. When you break one, you soon break the other.

Front view. Large, but acceptable.

Rear view. You can notice the battery compartment, and the broken body’s plastics

It happened on the first Motorola HS805 device I’ve had, and, somehow, I was able to replace it in the store. Two weeks later, careful as much as I could be, I’ve managed to break another such BT headset. I felt frustrated, but due to life’s course, put it aside, and moved on.

Two days ago, I’ve found it, and I’ve decided to invest the time, and try to replace it, or have it fixed. The store sent me to Motorola support and repair center, and there I was told that this BT was broken due to unreasonable usage (Me? I’ve used it in the most reasonable way! I replaced battery!), and they don’t have spare parts, so nothing can be done.

I was quite disappointed at that stage. I’ve expected it to be a known issue, and it seems I’m the only one who managed to break
his BT device… So I’ve looked on the net. I’ve found around 10 reviews (positive, I must say!), but they managed, somehow, to be in every damn site I’ve checked. These same reviews, starting with this guy saying it might not fit small ears, and ending with this one saying something about "it gets sometimes cut out in my car", on dozens of
sites. I better learn again how these sites work, but it sure decreased my confidence with the review systems on the net. Trying to add a review, everything seemed to work, but it never got into the system (two days has passed!). Weird.

I’ve linked some sites showing the same reviews. For you pleasure, and concern. It sure made me concerned.

Review1; review1 again; review2

There’s also one from Amazon, but that actually sums most of them up.

Long while and not much to talk about

Saturday, August 6th, 2005

I’ve been to our own little “August Penguin” just few days ago, and it wasn’t too good. I avoided the lectures (who needs secure rsync?!?), but sat and talked with friends about all kinda stuff. One of the advantages of such an event is that you get to meet lots of your linux-related friends, and you get to talk technicalities with whoever you feel like doing it. It’s the place for it. You speak of funny geekish things you had, and people around you laugh. You tell your saddest stories (how the storage went berserk, for example), and people understand what you’re talking about. That’s the fun part. Other than that, it wasn’t that good. Wasn’t worth the long drive, and the day off (well, the day off was worth anyhow, for what it is).

I didn’t do much since. I’ve refused a job offer, as the salary was too low, and the driving was too long, and I will be interviewed for AIX clusters soon, which I don’t like anyhow…

Our client, the one with the hosting server, got hacked. It was a minor thing, and some defacement, but he detected it a day too late, and we could not dig it out of the logs. Not only, we found out the hacker run sendmail, on some arbitrary port, and we just removed it, and tried to locate the hole in though he got in. It’s something with PHP. It’s always something with PHP. I have no idea as to what, especially when there are around 400 hosted sites on this server. We’ll know better in the future, when it happens again.

A server of a friend of mine was highly loaded in the last few days. It appears Apache was killing his machine. His Apache is supposed, in 90% of the cases, to respond once, and close the connection. I’ve helped him, and removed the keepalive, and although it decreased the number of running servers dramatically, there were few instances of apache still consuming high CPU. Using extended server status, and top, I was able to notice the high load was due to some specific PHP script, which got to wait for 10 minutes, each time. I was able to point out at that script, and he checked it, and found a bug. Small PHP bugs get to increase load on servers dramatically. One should be aware of it. Anyone can write PHP scripts, but only some can write it with performance in mind. Mind you.