Development [LWN.net]

A strategy for dealing with a flaky computer

By Forrest Cook
October 14, 2009

Your author has just made it through some rough times in the world of computer reliability and thought he should share his experiences with the LWN readers. It all started a few months back when his primary desktop machine, a 3+ year old Lini box with a 3 Ghz Athlon 64 processor (reviewed here) started to act up.

The machine had been running without too many problems for most of its life, it had served well as a test platform for a variety of hardware configurations and had run many versions of the Ubuntu distribution with much success. A few months back, the machine was upgraded from Ubuntu 8.10 "Intrepid Ibex" to Ubuntu 9.04 "Jaunty Jackalope". There were a few initial and easily solved issues with a runaway chippcardd4 process, the upgrade appeared to work. Except, the machine became noticeably sluggish and the problem was difficult to locate. The top command did not reveal any load problems, /var/log/messages had nothing interesting to show, nor did the dmesg command. The machine seemed slower, but it worked well enough that your author decided to live with it until Ubuntu 9.10 "Karmic Koala" was released.

After several months of using Jaunty Jackalope, things got really strange. While pounding on Firefox, the most used application, the machine's user interface would become locked up. The mouse still moved on the screen, but all keyboard input was frozen. It was still possible to connect to the machine from a remote host via ssh. Doing so and running top showed that the Xorg process was eating up the CPU. The /var/log/Xorg.0.log file revealed nothing. Luckily, it was possible to reboot the machine and avoid a risky crash and a time consuming 1TB fsck operation. This mode of operation continued for a while and your author really started to appreciate Firefox's ability to recover the previous session. It seemed like an opportune time to make a fresh backup of the machine to a remote host, something your author does every 2 weeks or so. That effort succeeded.

At this point, your author started suspecting flaky hardware. The keyboard seemed like a likely, if unusual culprit, it was swapped with no change. The video card was swapped out for a different model to force a change in the video card driver, still no improvement. It had been over a year since the CPU fan had been cleaned, this time, the fan was removed and "detailed" with a damp toothbrush. The CPU was cleaned and new heat sink grease was applied. The dust was blown off of the motherboard and power supply. The memory and I/O cards were cleaned and reseated. The BIOS was checked for proper power supply voltages. Memtest86 was run and the memory checked ok. Still, the machine flaked out under normal use.

At this point, your author was very glad that he had previously set up an old Athlon 1700 machine as an alternate work platform. That machine was used to perform the weekly LWN email and article processing tasks, it was slow, but slow is better than randomly flaky.

Around this time, the beta release of Ubuntu 9.10 "Karmic Koala" was announced. A copy of the beta iso was downloaded on the reliable machine and a disk was burned (and verified). An older 500 GB SATA disk was swapped into the machine and the new operating system was installed on that disk. The old 1TB SATA disk was installed as the secondary drive and all of the data (/home, your author's 250GB music collection and a few /etc config files) were copied to the 500GB drive. The Karmic Koala beta installation was a success, suddenly the Lini became much faster and it has been used heavily for over a week with no lockups. So far, the only problem with the new system has involved some folder refresh issues and key mapping problems with Claws-Mail 3.7.2. Otherwise, the system has been solid.

Upgrading to a newer distribution version, whether as an emergency measure or as a planned job, is much easier to do if you have one or more spare system disk drives. The extra drive can be thought of as a bootable data cartridge. The secondary backup on the remote machine allowed your author to sleep at night and the functional but slow alternate machine allowed work to continue without much interruption. Finally, the old drive with the flaky system is still intact if further investigation is required, it will soon become the next backup target. Your author has always had a preference for full installs over upgrades. It may be superstitious to blame the Intrepid->Jaunty upgrade as the root of the problem with no real data to prove that theory, but a fresh install will always clean out the cobwebs from the old system.

Comments (19 posted)

System Applications

Database Software

Firebird 1.5.6 released

Version 1.5.6 of the Firebird DBMS has been announced. "The project is pleased to announce that Firebird 1.5.6 release kits are now available for all the supported main-line platforms (Win32, Linux i86 and MacOS-X/Darwin i86 and PPC). No further sub-releases are planned for the V.1.5.x series."

Comments (none posted)

MySQL Server 5.4.3-beta released

Version 5.4.3-beta of MySQL Server has been announced. "MySQL 5.4 is based on MySQL 5.1 but includes several high-impact changes to address scalability and performance issues in MySQL Server. These changes exploit advances in hardware and CPU design and enable better utilization of existing hardware. MySQL 5.4 currently has Beta status."