Today I finally found the time to update my home server from NetBSD 3.1_STABLE to 4.0_BETA2. As usual the update went off without major hitches. I booted the 4.0_BETA2 kernel into single user mode and was pleasantly surpised that it automatically found the root filesystem on the hardware RAID. NetBSD 3.x kernels required a hardwired root filesystem to boot on my server.
Updating the userland was the next task: I extracted the base distribution binary sets, used postinstall(8) to handle most of the necessary changes and fixed the rest (missing users and groups) manually. In addition I replaced all installed packages with 4.0_BETA2 binary packages built on another machine a few days ago. After updating the boot blocks I rebooted the machine again and allowed it proceed to multiuser mode uninterrupted.
Most services worked without problems. Sendmail didn’t start because I hadn’t updated its startup scripts, Postfix tried to start because I hadn’t disabled it. I corrected both issues and my server has been working fine ever since (which is admittedly only a few hours at the moment). So far I had not much time to play with the new features. But I managed at least to switch the /tmp filesystem from MFS to tmpfs.
My next home server improvement project is to replace Sendmail with Postfix. But that will require more time and studying of the Postfix book to complete the configuration files.
Last saturday my friend Markus finally found the time to update the hardware of his externally hosted server beaver.core.de. The update took place month later than initially planed because Markus has a very busy schedule and his hobby server is understandably not his highest priority.
The new hardware consists of an ASUS P4B266-SE motherboard, a Pentium 4 2.0A CPU and 1GB of DDR memory. While this would be step backward for mankind it was giant leap for that system. The previous hardware was built in the last century: an ASUS T2P4 motherboard, a AMD K6 III 400MHz CPU and 192MB of PS/2 EDO memory.I was looking forward to the update because beaver is also the backup DNS and mail server for my domains. I help Markus with the system administration and if you are used to the speed of a machine driven by a Pentium M CPU it is not much fun doing similar things on such old hardware. Besides the slow CPU the lack of main memory really hurt the overall performance. The machine spent a lot of time paging out memory especially while SpamAssassin was processing e-mails. The high disk throughput caused by this was probably the reason that one IDE harddisks died earlier this year. And the other harddisk has hung once recently which indicates a forthcoming failure. Fortunately beaver uses NetBSD’s software RAID driver RAIDframe which prevented a catastrophe in both cases.
The new hardware seemed to work very well at first. The machine came up fine and felt really fast. Yesterday I had another quick look and sent an e-mail to Markus that everything is allright. That was of course a bad idea because the universe doesn’t tolerate such open display of optimism. Five minutes later the machine froze for reasons which we haven’t figured out yet. It came back after being resetted manually this morning. But I guess that it wasn’t the last time that this has happened. 🙁
Elad Efrat, one of the most active NetBSD developers in the last few month, recently organized NetBSD’s first Bugathon. It was a successful effort involving NetBSD users and developers to reduce the number of open bugs report against the NetBSD operating system.
I tried to contribute myself but was fighting with adverse conditions when I finally had some time. I nevertheless managed to take a look at two bugs:
- PR bin/24583: /etc/security doesn’t parse /etc/exports correctly
- PR bin/28492: more(1) doesn’t handle NUL data blocks well
The first PR was easily solved because Jukka Salmi already supplied a fine working patch. I only applied it on two of my systems (NetBSD 3.1_RC3 and 4.0_BETA), checked my e-mail the next morning to make sure that the nightly scripts worked fine and committed it.
I’m still working on the second PR because I cannot reproduce it so far.