HDD Preclear Shuffle in New Build

[Sol_Brotha]

So, I've finally put together a new build, and am excited to try out this unRaid thing.

Hardware is all connected, Memtest cleared two passes with no problem, on to Preclear!

My two HDDs are both refurbs, so I'd like to give them a thorough checkout, say, 20 passes of preclear.

However, I also want to get started playing with unRaid asap.

 

Here is my plan.

• Start HDD A and B on parallel preclears each of one pass.
  
• HDD B finished?  Assign it to Disc 1 and start playing around with unRaid (no critical data assigned to unRaid yet).
  
• HDD A finished?  Start it on more preclear passes (I read 20 cycles is the recommended burn in).
  
• HDD A finished again, in what, 15 days?  Assign it as Parity Disc.
  
• Let unRaid populate HDD A sync parity.
  
• Parity synched?  Pull HDD B off of the array, leaving DISC 1 in (emulated?) mode.
  
• Run HDD B through many passes of preclear
  
• HDD B finished its burn-in?  Re-assign it as DISC 1
  
• Sync Parity Re-build DISC 1 on HDD B, array is finally in protected mode
  

 

This should allow me to burn-in both discs without suffering another two weeks downtime. 

Does this seem correct?

 

Edit: corrected language regarding parity sync and disc rebuild

[trurl]

I've never heard of anyone doing that many preclear cycles. Doubt there is anything to be gained by more than 3. I think SMART reports will tell the story more than excessive number of preclear cycles. Disks don't really get better with "burn-in", all you really get is weeding out infant mortality. You don't say how big the drives are but I usually expect about 10 hours per TB for preclear.

 

I have a few nits to pick with some of your terminology.

• HDD A finished again, in what, 15 days?  Assign it as Parity Disc.
  
• Let unRaid populate HDD A.
   
  ...
   
   
  
• HDD B finished its burn-in?  Re-assign it as DISC 1
  
• Sync Parity, array is finally in protected mode
  

I suppose you could use the verb populate on the parity drive, but just be aware that parity doesn't really have any files on it.

 

Parity sync usually means building parity. You would be rebuilding a data disk.

 

I guess you could do what you say, though with a lot fewer preclear cycles, if you just wanted to see what building parity and rebuilding a data disk were like. What exactly did you have in mind for your server? Just NAS? Dockers? VMs? With that video card I guess you are probably going for VMs. You are probably going to want SSD cache. Cache disks are simpler to manage (unRAID takes care of it all) than drives mounted outside the array.

[Sol_Brotha]

Thanks for the input!

 

I've never heard of anyone doing that many preclear cycles. Doubt there is anything to be gained by more than 3.

The disks are refurbished, and if they throw a problem, I'd really appreciate it happening in the first 90 days.  I read the following from the wiki:

Preclear Disk...

    Why would you want to Preclear a drive?...

    This utility can also be used to burn-in a disk. If the disk can go through a full cycle without errors, it is less likely to fail shortly after being assigned as part of the unRAID array. You may request this script to perform multiple read/write/read cycles of the entire disk. As many as 20 cycles may be requested at a time...

20 cycles of pre-read/write/post-read would exercise a 1.5TB drive for over 14 days, a decent "burn in" period of time.

 

You don't say how big the drives are but I usually expect about 10 hours per TB for preclear.

2TB each.  I planned to do one pass initially to time the procedure before jumping in to multiple cycles.

 

I have a few nits to pick with some of your terminology.

Appreciated.  How is this:

• ...
  
• Assign HDD A as Parity Disc.  Let unRaid synch parity.
  
• ...
  
• Use parity to reconstruct Disc 1 on HDD B
  

 

What exactly did you have in mind for your server?... You are probably going to want SSD cache. Cache disks are simpler to manage (unRAID takes care of it all) than drives mounted outside the array.

I've been wanting to build a HTPC for ~4 years, a new gaming PC for ~12 (family life gave me other priorities), and a home file server for ~2.  I don't have the free cash to build out all three (see other priorities), so I figured build just the one, and let the magic of virtualization take over.  I intend to assign the 500GB SSD as cache.  My signature has it unasigned as I have yet to visit the unRaid configuration menu for the first time.  That's also why the box is still named 'Tower'.

[garycase]

As noted, 20 passes is FAR too much.    Do 2 or 3 ... on BOTH drives; and then just set up your array and use it.

 

The SSD can be assigned as a cache at any time -- although quite frankly with only two drives assigned to the array you don't need a cache, as the write performance will be FAR better with 2 drives assigned than it normally is with larger arrays.  When there are only 2 assigned drives, UnRAID recognizes this special case and effectively treats it as a RAID-1 array, so writes aren't slowed down by the usual 4-I/O's required to maintain parity with larger arrays.  [You may still want to assign it as storage for Dockers, etc. -- but there's no need to enable caching for your user shares until you add more drives.]

 

 

[trurl]

Yes, the cache drive often gets used for quite a bit more than just caching user shares these days. I never cache any user shares.

 

The advantage of using cache or a drive outside the array for apps (dockers, VMs), is that the parity drive is not involved when the apps need to write to their storage, so the writes are not slowed down by parity updating, and the parity disk does not need to spin up.

 

The advantage of using the cache drive for this instead of another drive that you have mounted outside the array, is unRAID already manages the cache drive for you. If you mount some other drive outside the array you will have to manage it yourself.

[Sol_Brotha]

Thanks for the suggestions.  I do intend to have unRAID manage the SSD as a cache drive without actually allowing user shares to cache.  It is good to know that (a two disk) array performance will not be impacted by this.  Once the array is up and running, I'll have 2TB of bulk storage and 500MB to play with for VMs.

 

Right now this feels like tons of space.  Any bets as to how long before I want to cram a dozen HDDs in there?

 

Preclear has been happily chugging away for ~20 hours so far.  Should be done soon.

I'll check the results and save them somewhere persistent on the flash drive.

Then, I'll need to shut down the system and muck about with cooling. 

 

The HDDs are reporting ~42C during preclear which has me nervous (isn't >45C dangerous?).  I have a 140mm fan venting air in their chamber, but it is speed limited for noise purposes.  I'll remove the limiter and see if that fixes things.  Also, it's a new build, so I should check that all the fans are blowing the right direction 

[itimpi]

The HDDs are reporting ~42C during preclear which has me nervous (isn't >45C dangerous?).  I have a 140mm fan venting air in their chamber, but it is speed limited for noise purposes.  I'll remove the limiter and see if that fixes things.  Also, it's a new build, so I should check that all the fans are blowing the right direction 

That is not all excessive.  Virtually all hard disks are rated to run at up to something like 65-70C without problems.  However many users like to keep the temperature as low as possible.

 

The actual temperature you get seem to depend a lot on the exact system, how it is ventilated, and also the make/model of disks that are used.  I would suggest that you watch what temperature the disks get up to when under a typical heavy load (such as the pre-clear or a parity check), and set the warning level to slightly above that so that you do not get warnings in normal use. Then set the critical level to something like 5C higher.  This will pick up things like fans failing so that the cases ventilation is compromised.

[garycase]

As noted, 42 isn't a bad temp -- and this is after many hours of continuous use of the drives.    They'll likely be notably cooler in normal operation.  The only time they'll likely get back to that level is during parity checks, which are also hours-long continuous usage operations.

 

[Sol_Brotha]

First round of pre-clear found no errors.  Yippie!

 

I investigated the cooling setup and did find room for minor improvement.

I changed the routing of the 140mm vent fan immediately adjacent to the HDD cage so that it blocks less airflow.

 

Nominal benefit was also to be had by remembering to plug it in. 

 

New batch of preclears running on HDD A using the web front-end. 32C during pre-read.