Downgraded cache drive, now it isn't being used during transfers

[Rajahal]

I was in the tinkering mood over the weekend and I decided to downgrade my cache drive from a 3.5" 320 GB Seagate to a 2.5" 200 GB Fujitsu that I had laying around.  This drive is known to be somewhat bad (gets about a 40% fitness rating on the SMART test), but it does still work, and I would like to use it until it completely dies.  I figured using it as a cache drive would be appropriate since at worst I would lose the last day's worth of transferred downloads (which would still be located on my torrent box), plus it would clear up my previously used cache drive to be a data drive.  The swap went smoothly, I just unassigned the previous cache drive, assigned it as a data drive, then assigned the new drive into the cache slot.  unRAID recognized it, cleared it, and formatted it with no complaints.  None of this is my question, but I welcome comments on my logic of using an unreliable drive as a cache drive.

 

Here's my real question - now that the new, smaller cache drive is installed, why isn't it being used when I copy files to my user shares?  Keep in mind that the capacity of the drive is now diminished by 120 GB....

 

My thoughts: I noticed that on the 'settings' page the 'min free space' number for the cache drive didn't change after I swapped them.  It remained at 2000000 or something like that (200 GB?).  I figured that it was set too large for the new cache drive (since it only has ~195 GB usable space) and tried setting it to 1000 (1 MB?).  Still didn't work.  So I tried the other end of the spectrum and changed it to 100000000 (1 TB, 10 TB?).  Still nothing.  I've got a hunch that my problem revolves around this number, but I don't know what I should set it to, or how to interpret what it means.

 

Note: This new cache drive is not the one reporting SMART errors that I'm asking about in this thread.  I know this absolutely because I got those SMART errors before I ever installed this drive.

[RobJ]

This is probably too simple, but I'll ask anyway, have you rebooted?  Some security and share settings seem to take effect on the next boot.

 

Concerning the Cache floor setting, in the absence of any stated units, I would guess it is in bytes, but a default of 2 million bytes seems WAY too small for drives which are so commonly used for video recordings.  The next most common unRAID unit is the 1K block, so perhaps it is in those, which would make it about 2 gigabytes, still too small as a floor for videos.  Depending on whether you are storing HD, the floor should be anywhere from 8GB to 30GB, perhaps even bigger for Blu-Ray's.

 

As to using 'unreliable' drives, it's a personal choice, with some strong opinions on one side.  I personally would use the drive as you have, but I know that some others would definitely not.

 

Note: the following are just my thoughts, about drives that appear imperfect, and I in no way meant to 'pick on you' here.  You got me started though!

 

As to that drive, are you sure that it is unreliable?  And how did you obtain that "40% fitness rating"?  I personally don't trust those SMART programs that attempt to make a fitness or lifetime or time-to-live type judgment, as there is no known reliable way of determining such.  There are many kinds of damage, and the drives are designed to work around many of them, and keep going.  There are often defects in their manufacture, but once the drive has lasted a month or so, especially if it had our Preclear testing done, then the drive has mapped around the defects and calibrated itself appropriately, and should last a long time.  If the drive is dropped or suffers a severe electrical spiking, there may be new damage reported, but again, it will recalibrate if necessary and remap any damaged sectors, and then it should be good to go for more years.  The fact that a SMART report shows damage does not in itself mean the drive is bad, but that the drive has dealt with damage and defects.  That is an important distinction.  Evidence of damage in a SMART report is not just letting us know of defects, but letting us know that it detected and dealt with them as best it could.  What IS bad is if consecutive SMART reports show a TREND of damage or failure, a continuously worsening state.  It really takes a series of SMART reports to show us that a drive is failing, by showing us negative trends.  That's why BubbaQ's SmartHistory addon is a valuable preventive maintenance tool, highly recommended.

[Rajahal]

Thanks for the insightful reply, RobJ.  To answer your question, I have restarted the server several times, but there was no change in the cache drive behavior.

 

Do you use a cache drive?  If so, what is your 'min free space' number set to?  I would be happy just to copy yours.  If it still doesn't work, then I must be having more serious problems.

 

Regarding the reliability of the drive, here's the full story: A few weeks ago I received a laptop to repair (I run a small computer repair and data recovery business as a side job).  The owner told me that it stopped booting correctly and asked me to fix it.  As is generally the case with modern laptops, the hard drive was very easy to access and remove from the laptop (just 2-4 screws), so my first step was to remove the hard drive, hook it up (internally) to my desktop, and run GetDataBack NTFS on it to recover his data (that program is slow, but amazing, by the way).  This was successful, but it took over 36 hours for the full process to complete.  This tipped me off that there might be something wrong with the drive itself.  I already had SpeedFan installed on my desktop (another great piece of software), so I used its built-in SMART analysis function to run short and long SMART tests on the drive (the short test takes just a few minutes, and the long test takes many, many hours).  SpeedFan does spit out a detailed SMART analysis of the drive (which I can post, if you are interested to see it), but it also sums up the SMART test into two easy categories: fitness and performance.  This drive got a 40% (out of 100%) as its fitness rating, and, if memory serves, about a 60% as its performance rating.  Most of my known good drives get 90%+ on both ratings, though I have never seen a full 100% rating on any drive, even on brand new drives right out of the box.  I informed the customer of this, and he decided he wanted to replace and upgrade the drive.  I bought him a brand new 500 GB drive, plopped it in, and the laptop ran just fine after that.  I kept his defective drive for a few weeks just in case there was some file he was still missing.  There wasn't, so I adopted the drive for myself.  At first I tried placing it in a 2.5" enclosure, but Windows wouldn't recognize it (which may be a problem with the enclosure, not the drive).  I then had the idea to use it as a cache drive in my unRAID server, so I went for it, as described in my original post.  I had mixed success, since unRAID recognized, cleared, and formatted the drive without error, however, now won't use the drive for anything.

 

I just had a thought.  Tonight when I get home I'll try copying a file directly to the cache share to see if the drive works at all.  If it works for that function, then it seems I should be able to get it to work as a cache drive for my user shares, right?  If it doesn't work for that function, then maybe unRAID is purposefully avoiding it because it is bad or something like that.

[RobJ]

Do you use a cache drive?  If so, what is your 'min free space' number set to?  I would be happy just to copy yours.  If it still doesn't work, then I must be having more serious problems.

No, sorry.  I have never even turned on User Shares.

 

(I run a small computer repair and data recovery business as a side job)

That is what I have wanted to do, but had to give up each time.  I love solving problems, fixing computers, helping people, but have found I'm helpless and useless at the mundane stuff, hate charging people, don't like paperwork, terrible at calling people back, etc.  If I *ever* find a friend or partner or sponsor, I'd love to do more of that.

 

I strongly agree with the excellency of GetDataBack and SpeedFan, great tools.  I haven't used SpeedFan in at least a year (crashes because of one damaged drive I have), but I do remember it had very good SMART reporting, and especially the detailed linked data, perhaps the best in the business.  And I do remember now that 'fitness rating'.  Although tempting, a nice graphical representation, I wouldn't trust it that much, even with the expertise evidently behind it, as evidenced by the quality of their reporting.

[RobJ]

If so, what is your 'min free space' number set to?

 

In the unRAID v4.5-beta5 Release note is an explanation of the "Min free space" setting, in the paragraph that begins with "Along with split0 feature".