knew
For future reference, this is the knew server. It runs a few jails, including Bacula regression testing services. The filesytems:
Disks
slocum
This post has been replaced by a newer one. For future reference, this is the slocum server, which I use for various jails and services:
ZFS: raidz2 with 8 or 10 disks?
The traditional thoughts on ZFS are to have arrays which are a power-of-2 when it comes to number of disks. That means 2, 4, 8, etc. But for raidz1, that would be 3, 5, 9. For my choice of raidz2, it would be 4, 6, and 10. In the system I’m building up now, I have an 8-port LSI SATA/SAS 9211-8i 6Gb/s card. I can add two more HDD by attaching them to
ZFS: raidz2 with 8 or 10 disks? Read More »
smartctl: FAILING_NOW
I noticed this in my logs today: May 22 14:04:39 bast smartd[1577]: Device: /dev/ad0, Failed SMART usage Attribute: 9 Power_On_Hours. Oh. Really? I looked at the output of smartctl and found: This HDD is part of a gmirror: $ gmirror status Name Status Components mirror/gm0 COMPLETE ad0 ad2 How’s ad2 doing? That seems OK. Regardless of the status, I can do nothing for a week. I’m out of town. I do have several
smartctl: FAILING_NOW Read More »
The Bacula Tutorial jail server
One of the challenges of providing hands-on demonstrations is giving everyone their own sandbox to play in. I don’t want people to spend time on installing software. I want people to learn about the software in question, specifically Bacula. With this in mind, I’ve been building up a solution based on FreeBSD 9.1, ZFS, and jails. My solution is pretty nifty, but I don’t think it’s anything special. The key is simplicity. The
The Bacula Tutorial jail server Read More »
Oh oh, a sector problem
I woke up to this today. Complicating the matter: this server is destined to be transported tomorrow morning. This is FreeBSD 8.2 I guess the good news is: But wait, is that gpt/disk04-live really /dev/ada5? Yes. Yes it is: After a question from Peter Wemm on Twitter, I added: Full smartctl output follows:
Oh oh, a sector problem Read More »
same HDD, different server, +50% performance hit
As you’ll no doubt know, I’ve recently bought two new computer, and built them up. I’ve settled on the Toshiba 3TB DT01ACA300 HDD as my drive of choice for these systems. I ran a few tests recently and found some interesting results. I haven’t published them here yet. My main test is a database load. I just take an 8GB text file and create a PostgreSQL database from it. The file in question
same HDD, different server, +50% performance hit Read More »
This harddrive is getting on in life
While on a recent trip, logcheck started telling me about a problem: Mar 19 07:26:55 bast smartd[1573]: Device: /dev/ad0, Failed SMART usage Attribute: 9 Power_On_Hours. Looking at the full output: On line 59, you’ll see that this drive has been in use for 7.4 years. Other than that, it seems to be OK. This link seems to indicate there is nothing to worry about. This system dates from at least October 2005. It
This harddrive is getting on in life Read More »
Oops, I deleted my boot partition
About a week ago, I was doing some HDD benchmark tests and I accidentally specified the boot drive as the source for a test. Oops. Everything was OK. Until I rebooted. It would not boot. I figured I knew what happend. I’d munged my boot code. And I knew how to fix it. However, I didn’t realize that I’d also deleted the boot partition. I asked on the FreeBSD dev IRC channel. db
Oops, I deleted my boot partition Read More »
Automating that regression testing
I’ve been testing drives lately. Mostly hard disks, but also one solid state drive. The test itself is automated, but manually started. That’s because it requires a reboot between each run. That restricts testing time to certain hours of the day. I’d like to automate this. Today I had what may be a clever solution: Use the @reboot feature of crontab(5) to start a script after reboot In that script, cd to a
Automating that regression testing Read More »