On Sep 3, 2010, at 10:10 AM, Charles Scott wrote:

> Gary:
> 
> Well, it's kind of a Tim "The Tool Man" Taylor tendency I have. I did, 
> however look at the Icom specs for the gateway system and it says 2.4 GHz and 
> 512 MB, so I'm not even an order of magnitude over that, which would be the 
> "Tim" thing to do. I also considered fan-less 12V computer boards, but for 
> what I paid for these systems, I couldn't buy a new one of those boards.
> 
> The nice thing about this system is that it has redundant supplies, BIOS, 
> drive array, fans, and so on so it shouldn't go down with common failures. It 
> also has integrated lights-out management, so I can can talk to it over the 
> network even when it's shut down to restart it, reboot, or diagnose problems 
> remotely. The other nice thing is that they show up regularly on E-Bay at 
> prices less than a cheap PC at Walmart. (It's a DL385 G1.)
> 
> If nothing else, I could chew up some CPU cycles doing s...@home or some such 
> thing, but the IRLP computer did end up with a bunch of things running on it 
> also, and I suspect this will be the same.
> 
> I even thought of installing VMWare on it and running both the D-Star gateway 
> and IRLP in separate virtual machines, but I'd have to get more memory in it 
> to do that well.
> 
> Chuck

Think about doing RAID1 and having two disks in it if it's inaccessible for 1/2 
of the year.  

Disclaimer: I did this with W0CDS, which lives on top of a very high mountain 
-- and it still bit me in the hindquarters.  Linux Software RAID1 isn't 100% 
ready-for-prime-time, sadly, after all of these years.

The machine lost a drive, and instead of just chugging along, it started 
throwing I/O errors for all commands. 

Luckily, the RAID was working, it just never "detected" the failed disk.  A 
power off/power on reboot cleared that problem and it came right back online 
with a single disk and one in a failed state in /proc/mdstat -- so that leads 
to item #2...

Get a way to remotely REBOOT your system... be it a transistor switch on a 
co-located analog repeater controller, a remote power on/off device like a 
managed power strip, whatever works that you trust and can access when the box 
is down.  

That would have saved someone a trip to the mountain. 

But he went, we proved the machine would run on one dead drive and one live 
drive, and then he yanked the dead one to bring it down to get a replacement.

Which leads to item #3... 

Since Linux Software RAID can work with, but really really really likes drives 
of the exact same CHS layout and size... get a couple of spares.  Drive 
technology is still changing so fast, that by the time you need it, that model 
will be hard to find.  Drives are cheap, keep spares if you're using RAID, or 
be prepared to backup/rebuild the system from scratch with two new drives when 
one finally fails.

The reason for the drive failure we suspect is two-fold... high altitude (heat, 
less air between the spinning disk and the "flying" head, etc) and really bad 
power up there.  Lightning wreaks havok with everything up there every summer, 
and apparently this drive died too soon into its usual life-span because of all 
the power hits.  Even once we had a UPS inline, the "stuff' that comes in on 
the power lines up there is just utter trash all summer long.

It's just a tough environment for PCs.  If you're building from scratch and 
don't mind the eventual performance hit and need to do a "secure wipe" and 
reload once in a while, the modern Solid State Drives are a good choice for a 
difficult site, I think.  But their internal fragmentation problems and limits 
are becoming well-documented, and that "secure wipe" to get them to go rewrite 
every bit of the flash and reset the controller that's managing the flash, is 
important for most brands.  Some good reviews of cheap vs. "server grade" SSDs 
are starting to show up on the web in droves now, whereas for a few years 
there, the testing and performance numbers just weren't available.  

I'd say it's a toss-up between spinning platter and SSD, when you factor in 
price.  Cheaper than owning two SSD's is owning four mid-sized spinning platter 
technology drives, so you'll have to decide if you want to pay the premium and 
be an early-adopter, so to speak.

--
Nate Duehr
[email protected]

facebook.com/denverpilot
twitter.com/denverpilot

Reply via email to