Linux Help with RAID 5 on Xubuntu file server

Hi,

After following Glider's rather excellent server guides last year (and continually asking him questions ever since ) I have a very solid and nippy little fileserver.

The touble lies in the word, "little". I only have an 80GB drive in there (it was initially just 'an experiment' you see) and desperately need more space.

The old mobo in there has 3 spare SATA ports, so I was considering adding an additional 3 x 500GiB drives in software RAID 5 (whilst leaving the boot drive as it is). As I understand it, that will give me 1TB of 'useable' storage on the 3 drives (with 500GB reserved for parity).

I found this guide which makes it all look fairly straight forward and I'm sure I'll be able to follow it.

However, I do have some questions (some may make you laugh) about the procedure and software RAID in general:

1. How would I know if a drive had failed (and hence needed replacing & rebuilding) in a headless fileserver?
2. How would I identify which drive physically needed replacing when a problem occured?
3. Is a 3.06Ghz pentium 3 with 1GB of RAM gonna hack it?

 

1 and 2 i can't help with, but for #3, i've run software raid 5 of a gen 1 mini-itx board before (800mhz via CPU with 512mb PC133 ram, somewhat less performance than a 500mhz pentium3) with no noticeable reduction on speed.

Was limited to a 100meg LAN though, i'm guessing gigabit will be more demanding, but for 100meg you have no worries at all.

never saw my main PC go much above 5% cpu usage with a cheapish mostly-software raid5 card either... max 5% cpu usage under extreme HDD testing on a 6000+ X2 CPU...
I feel the CPU-hit of software/part-software raid is massively overblown...

 

Thanks, I agree to be honest.

I think particularly for a (domestic) file server, it can spare the necessary clock cycles for the parity calcs.

Unfortunately, it's the first 2 points that are making me hesitant to proceed.

 

You would either need to login remotely at regular intervals, get a status report sent to you (internal email?), or have a web-based health monitor (webmin?).

Any of the above.

More than enough power.

 

Must admit i've not had any experience with drives failing in raid arrays... but it generally is fairly obvious when a drive fails badly due to the noise ...

If you don't get horribly grinding noises/whatever from a drive if a drive in the array fails, you could boot an OS off a CD and do a quick test of each HDD manually one at a time...
It'd take time, but still should be fairly quick to spot a suspect drive

 

One of those (and much more possibility):
-Foward/alias your local mail (every accoutn normally have a mail box on linux, for cron stuff mainly) to a "real" mail box, or setup the internal one to availible outside of localhost.
-Do a cron job to check the state and send an email
-Put a 5$ LCD/LED display on the serial port of your file server to display drive health.
-Use monitoring software
-Find a circuit with a serial port and a transistor to light a LED for each drive using USBserial/COM1/LPT1. do something like
while [ `<comamnd to check drive> | grep State | grep -o "running"` ]; do echo 111111111111111111111111111111111111111 > /dev/ttyS0; done
to trigger the transistor, but it will eat a lot of cpu, so it is a bad idea, but a State LED is cool. You can also une a PIC chips to do it, but now it will be much harder. You can also use a buzzer (irriting sound things) instead of a LED if you dont see the file server. The command is mdadm if my memory is good, but i never used it.

 

Not if you do:

Code:

while [ '<command to check drive> | grep "State.* running"' ]; do
  echo 111111111111111111111111111111111111111 > /dev/ttyS0
  sleep 1
done

 

but then it will trigger off during 1 second, it need continuous input if it don't use some kind of microprosessor (PIC).

 

Ah, good point.