Geekly musings

[unixronin]

For starters, yesterday's DSL problems should be over for good.  I just received and installed the replacement DSL bridge/"modem".  As an added bonus, it's an upgrade, a Westell 6100 instead of a 2200.  It wouldn't auth to Verizon's servers at first, even with the reset password that was just set yesterday, so I had to call Verizon support to get it reset again; even then, that was unexpectedly pleasant, as I chanced to get a tech who was at least somewhat knowledgeable about -- or at least interested in -- Linux and *BSD, so we chatted about those for a few.

Minor bonus:  the 6100 is black, instead of the ugly generic-computer putty color of the 2200, and is laid out such that it should lent itself very well to mounting on the side of the rack once I figure out a few minor implementation details.  Some self-adhesive ferrite-rubber magnetic sheet may work.

In other news, I was recently given an external HP Ultrium LTO1 100/200GB tape drive, but had problems getting it working with Bacula.  The first step was to swap out babylon5's disk controllers; it's been running for a long time on an old Mylex BT958C and an equally old Adaptec AHA2940.  The Mylex was a brilliant controller in its day, with very low latency and fast as stink, but it's long since been surpassed.  Both controllers got replaced with Adaptec 29160s, which are LVD controllers, still single-channel but with the capability to segment their bus into high, middle and low speed segments instead of limiting all devices to the speed of the slowest device on the bus.  After I did that, connecting up the drive was simple, and my god, the thing sync-negotiates at 80 megabits per second -- that's faster than my disks.  However, bacula's file daemon would crash every time it tried to access the LTO1.  I couldn't figure out a reasun, until suddenly I realized "Doh!  TERMINATION!"

Of course, I couldn't find my spare active SCSI terminators.  But I found one on eBay for $1.99.  Predictably, after buying it, but before it arrived, yesterday I found my spare terminators.  Attach an active terminator, and suddenly Bacula is happy with the LTO1.  After a number of dry runs last night with considerable tweaking along the way, I finally dumped my entire Bacula catalog this morning, cleared and recreated my tape pools, relabelled all of my tapes (five 30GB VXA1 tapes and four 100GB LTO1 tapes), enabled concurrency on the LTO1 drive, and ran full backups of everything.  This thing really steams; Bacula dumped 168GB to tape from eight clients in five hours, averaging close to 10MB per second aggregate transfer rate most of the time, and achieving about 30% hardware compression.

Sweet as.  I now have sufficient capacity to turn on regular scheduled backups again.

Comments

[robbat2.livejournal.com]

from the pair of LTO1 drives that we had at my previous job, we found that the hardware compression really hurt the actual speed, and we more than doubled speed by just using host-side gzip -1 compression, and turning off all hardware compression.

[unixronin.livejournal.com]

In this case, that'd be Bad. I'm already CPU-bound on the catalog updates. That'd slow my overall throughput way down, I think.