Backup UnRAID server specs

[JimPhreak]

In addition to the size difference, the older technology uses a lot more power than a new Haswell or Avoton board would;  uses more expensive and less efficient DDR2 instead of DDR3;  doesn't have a warranty;  etc.

 

... In short, the modest cost difference has tradeoffs that come with it -- personally, I wouldn't buy that board even if it was mini-ITX 

 

Yea I completely agree with all of that and was never considering it.  But the fact that it's not Mini-ITX eliminated it from being an option anyway so I didn't feel the need to explain from there .

[pkn]

I see. Server which will be powered up at most once a week, especially needs modern energy saving technology.  But that's OK, I too like new and shiny more than old and dusty...  I was just little confused by talks 'bout budget.

[JimPhreak]

I see. Server which will be powered up at most once a week, especially needs modern energy saving technology.  But that's OK, I too like new and shiny more than old and dusty...  I was just little confused by talks 'bout budget.

 

No worries.  I definitely had a specific budget in mind when I started this thread but as you can tell as this thread has evolved, that budget has gone up by about $300  .

[JimPhreak]

Finally have my UnRAID server up.  UnRAID installed, license installed, yet having an issue with drives not being detected.

 

I have 6 drives connected to an IBM M1015.  Only half the drives are being detected at a time.  If I have both SAS cables plugged into the M1015, only the first 3 drives are detected.  If I unplug port 0 and leave just port 1 plugged in, the other 2 drives are detected.  Very strange.

 

When I got the M1015 I flashed it to IT mode btw.  Not sure why only half the drives are showing up with both SAS cables plugged in.

[garycase]

The drive detection issue sounds like a possible power problem.

 

Do they all detect okay if you reboot without powering down?  [if you don't have adequate spinup current capability this will sometimes resolve that, since the drives are already spinning when you do the reboot ... although it's not really a good "fix", since that means a "Spinup" command or parity check would likely cause issues when all the drives tried to spin up at once]

 

Looking back at the configuration, you indicated you were going to buy a 350w Seasonic 1U power supply.    350w is PLENTY of power for this system ... HOWEVER -- I looked up the specs for the power supply you indicated you might buy, and it has 2 12v rails.    If your drives are all connected to the same rail, that may be what's causing this issue.  You may need to do a bit of re-cabling to ensure the load is split between the 2 rails.

 

[JimPhreak]

The drive detection issue sounds like a possible power problem.

 

Do they all detect okay if you reboot without powering down?  [if you don't have adequate spinup current capability this will sometimes resolve that, since the drives are already spinning when you do the reboot ... although it's not really a good "fix", since that means a "Spinup" command or parity check would likely cause issues when all the drives tried to spin up at once]

 

Looking back at the configuration, you indicated you were going to buy a 350w Seasonic 1U power supply.    350w is PLENTY of power for this system ... HOWEVER -- I looked up the specs for the power supply you indicated you might buy, and it has 2 12v rails.    If your drives are all connected to the same rail, that may be what's causing this issue.  You may need to do a bit of re-cabling to ensure the load is split between the 2 rails.

 

Power could be the issue, that makes sense.  My hard drive are plugged into a backplane that is powered by two 4-pin molex plugs and my PSU only has one cable that has 2 molex plugs on it.  So yes that means all the hard drives would be plugged into the same 12v rail.  I'll have to get a SATA > Molex adapter so I can use a one of the SATA power cables to power half the backplane.

[JimPhreak]

Well if it is a power issue, doesn't seem like this PSU will work for me.  When I tried using just one of the molex power cables (instead of the two on the same cable) and one SATA power cable (with SATA -> Molex adapter on it) and that doesn't change anything.

 

What's weirder is what I found though when I only power one of the two molex connectors for the SATA back-plane.  I thought originally that if I only plugged in one or the other, either 3 drives showed up or the other 2 drives showed up.  However when I see just the two drives, it's the same two drives that show up when I get 3 drives (with one missing of course).

 

Something strange is going on here.  More troubleshooting ahead.

[garycase]

Do you have another (preferable single rail) power supply you can try?    Don't need to put it in the case -- just set it beside it and run the cables inside.    That would let you confirm whether or not it's a power issue.

 

[JimPhreak]

Do you have another (preferable single rail) power supply you can try?    Don't need to put it in the case -- just set it beside it and run the cables inside.    That would let you confirm whether or not it's a power issue.

 

Just did that.  Tested half the backplane being powered by one PSU and the other half with the other PSU.  Also tested powering the full backplate with just the test PSU powering nothing but that while the other PSU powers my system.  Still getting the exact same result.  I'm going to need to recheck all my SAS -> Sata cables at this point I think.

[JimPhreak]

Wow it looks like 3 out of my 6 HDD's are bad.  One was bad right from the start (clicking sounds) so I knew I'd have to RMA that one.  But the other 2 aren't read by the system at all even if I put them into the same slots that the 3 good hard drives are being read from.

 

I even flashed my M1015 back to IR mode to see if it was something to do with the card.  It reads the 3 "good" drives just fine but not the other 2.

 

Man this sucks.  I was really hoping to get this server up by this weekend.  Grrrrrrrrr.

[garycase]

Wow !  That's a really bad infant mortality rate.

 

Of all the WD Reds I've bought, I've only had one shipment that had bad drives in it ... but that was clearly due to packaging issues (from Newegg).    When they've been well packed, I've never had a bad drive ... and they're all still going strong.

 

Were your drives packaged well when you received them?

 

[JimPhreak]

Wow !  That's a really bad infant mortality rate.

 

Of all the WD Reds I've bought, I've only had one shipment that had bad drives in it ... but that was clearly due to packaging issues (from Newegg).    When they've been well packed, I've never had a bad drive ... and they're all still going strong.

 

Were your drives packaged well when you received them?

 

Yes, they were all in individual boxes and packed inside a bigger box that has decent packing to ensure the boxes wouldn't move much...However now that I look at the big box, it does have a big gash/rip in it that would suggest it may have been dropped or something heavy was dropped on it.

 

Oh well, not really much I can do about it.  I guess this weekend I'm building my new pfSense box (luckily all those parts are here as well) while I wait for the new hard drives to arrive.

[JimPhreak]

How valuable is ECC RAM in an UnRAID setup?  I suspect it is important since the UnRAID file system has no checksums built into it.

 

The reason I'm asking is because I forgot I had one of these laying around and it would be pretty ideal for my backup server.  However I have some pretty specific use cases for my backup server that may may using non-ECC in my backup server and issue.

 

Even though my backup server is just that (a backup of my main UnRAID server), it also will house a local share for the PC(s) I have in that location.  I want to basically mirror both my UnRAID servers so that means I'll technically be backing up (one share) from the backup server to my main server for this one particular share (some personal data/photos, etc.).  With no checksums in UnRAID, how much of a risk am I at for data corruption during the mirroring process?

[garycase]

Given a choice, I'd always use a board that supports ECC;  but it's not a major issue if you don't have it.    Given that you already have that board, I'd just use it.

 

It does, of course, mean that you could indeed have some data corruption if you had a memory error during the read of the data.  This would probably be caught if you did a validation of the copy; but that depends on (a) how the validation is done; and (b) whether or not the exact same bit error occurs during the validation [in which case the error would not be caught.]

 

The safest way to protect against that is to keep checksums of your data and do periodic validations of those checksums.

 

 

 

[JimPhreak]

Given a choice, I'd always use a board that supports ECC;  but it's not a major issue if you don't have it.    Given that you already have that board, I'd just use it.

 

It does, of course, mean that you could indeed have some data corruption if you had a memory error during the read of the data.  This would probably be caught if you did a validation of the copy; but that depends on (a) how the validation is done; and (b) whether or not the exact same bit error occurs during the validation [in which case the error would not be caught.]

 

The safest way to protect against that is to keep checksums of your data and do periodic validations of those checksums.

 

Ok sounds good.  I'll look into setting up some sort of checksum after I get my backups configured.