Preclearing when your unRaid is full?

[gorbachev]

How do you guys preclear new drives for systems that are full?

 

My unRAID system is full, my case has no empty drive cages or SATA connectors, so I can't really use it.

 

I'm slowly replacing older drives with bigger, newer ones as I hit capacity. Preclearing used to be easy when I could just plug the new drive in an empty drive cage. No more.

[c3]

1) use another system, doesn't need to be much to just run preclear for 1,2 drives.

2) preclear on USB, much slower but often the drive comes as external.

[SSD]

I recommend users plan for an empty slot. This can be used for preclearing, but is important for some data recovery scenarios. This can even be an external setup what you pass a sata cable and power plug out the back of the server and run the drive on its back. Clearly you have to be very careful, but I've used this technique from time to time.for a temporary need. Just make sure it can't be disturbed by pets, kids, or vacuum wielding wives.

[tdallen]

Preclearing has two purposes, testing the drive and preparing it for the array if you want to add it empty.  I usually test new WD drives with WD Data Lifeguard in my desktop.  That's the lengthy part of the process, it takes days.  I'd feel comfortable rebuilding directly onto a drive that's been through a couple of passes with Data Lifeguard.

[SSD]

2 hours ago, tdallen said:

Preclearing has two purposes, testing the drive and preparing it for the array if you want to add it empty.  I usually test new WD drives with WD Data Lifeguard in my desktop.  That's the lengthy part of the process, it takes days.  I'd feel comfortable rebuilding directly onto a drive that's been through a couple of passes with Data Lifeguard.

 

Never used manufacturer drive testing tools. Always felt that SMART attributes are tuned to virtually NEVER show that the drive is failing. I remember a terrible drive where a normalized value was at 1 for reallocated sectors, and drive was unreadable, that didn't show "failing now"! Always felt that type of thinking (which I'm sure helps discourage  returns) would influence its drive testing suite. But I could be wrong. Have you found the tools to realistically highlight problems? I assume that the tools will only operate on drives from WD (or whichever manufacturer), so you'd have several different toolsets.

 

I am quite fond of preclear. @Joe L.put a lot of effort into it, (which I never thought her got the credit he deserved), and always think fondly of him when I use it.  I also enhanced it a little which helped me better appreciate the joys (or lack thereof) of shell script programming. (In fact my coding of the intermediate preclear statuses shows up in unRaid's GUI.)

 

Preclear has also proven very dependable at unearthing early failures. In fact, the once or twice I didn't preclear (very long ago), I had drive failures. Probably coincidental, but part of me thinks the gentle mostly sequential action of the preclear has a positive affect on the drive, similar to the break-in period of a new car. I have learned that data centers do not test drives, finding the exercise costly (someone had to do it and it takes s lot of time), and actually reduces the drive's useful life.

 

People here are fond of running drives through multiple preclear cycles, but I've always felt one was enough. When you think about the expected writes a drive would receive in its lifetime, writing every sector on a drive could easily represent choose to half of the drive's lifetime of writes. Doing 3 passes would raise that to 150% or more. Like a 100,000 mile test so the car won't fail in the next 25,000 miles (maybe not a good analogy but best one I can come up with). There was an old poll in the forums asking users about drive failures in different preclear cycles, and I remember that a very large percentage of failures were found on the first pass, with much diminishing returns after that.

 

But I've found everyone has their own preclearing ritual, and no one is going to change it based on others input because it had worked for them.

[tdallen]

I actually use a combination of WD Data Lifeguard to beat up the disk and CrystalDiskInfo to get the real SMART attributes since WD only presents thresholds.  My preference would be to use Preclear but like the OP I'm slot constrained so having an option to build confidence in the drive outside my server is nice.  I had a bad experience stringing eSATA cabling outside the server case, it generated a lot of UDMA CRC errors.

[SSD]

Data cable can't be too long.

 

I've also had luck booting unRaid from a usb stick in a random machine and running preclear. Once complete, can remove USB stick and it can boot from its hard disk as before. I have s small graveyard of old machines that I can use that have been put out of commission by my server! 

[gorbachev]

Good suggestions.

 

I think what I'm going to do with my rig is to replace the cache drive temporarily for preclearing the new disk(s). It seems like the path of least resistance to me.

 

For longer term I actually have two more SATA interfaces available, but no empty HDD cages. I'll have to get some SATA power cables for the PSU so that I can plug new drives in. I'll have to figure out a safe way to hold the drives during preclearing, but that's it.

[sonofdbn]

On 10/4/2017 at 11:19 PM, SSD said:

 

Never used manufacturer drive testing tools. Always felt that SMART attributes are tuned to virtually NEVER show that the drive is failing. I remember a terrible drive where a normalized value was at 1 for reallocated sectors, and drive was unreadable, that didn't show "failing now"! Always felt that type of thinking (which I'm sure helps discourage  returns) would influence its drive testing suite. But I could be wrong. Have you found the tools to realistically highlight problems? I assume that the tools will only operate on drives from WD (or whichever manufacturer), so you'd have several different toolsets.

 

I've used the WD Data Lifeguard tool with both HGST and Seagate drives without any problems. (Since HGST is, I believe, owned by WD, perhaps that's not surprising.)