How to preclear drives when I have no bays left?

[squarefrog]

I have an HP N40L, with 6 drives in (the maximum I can fit in). One of the drives is 250GB and failing. I've ordered a 3TB replacement, which I'd like to preclear before swapping out the 250GB.

 

How can I achieve this without significant downtime on my unRAID setup? Having read the Configuration Tutorial, it seems the recommended way is to use the Preclear Plugin, but I can't attach the drive to my NAS.

[itimpi]

I have an HP N40L, with 6 drives in (the maximum I can fit in). One of the drives is 250GB and failing. I've ordered a 3TB replacement, which I'd like to preclear before swapping out the 250GB.

 

How can I achieve this without significant downtime on my unRAID setup? Having read the Configuration Tutorial, it seems the recommended way is to use the Preclear Plugin, but I can't attach the drive to my NAS.

One simple option is to plug it in via USB assuming you have a suitable adapter and run the preclear that way.  If it is USB then it is of comparable speed to doing it via SATA, although USB 2 is significantly slower.

 

Another option is to stop the array; unassign the 250GB drive; and start the array with that drive missing.  unRAID will now 'emulate' that drive using the remaining drives plus parity.  That will free up a slot to plug in the new 3TB drive for pre-clearing purposes.  Once the preclear has finished and the drive checks out OK you can assign it in place of the missing drive and unRAID will rebuild the 250GB drive contents onto the new drive.

 

I assume you already have a 3TB (or larger) parity drive as parity can never be smaller than the largest data drive.

 

Also note that strictly speaking it is not necessary to preclear a drive which is going to be used for rebuild purposes.  Having said that since this is a new drive the preclear process is a good confidence check that the drive is OK.

 

In terms of downtime there are a few seconds the first time when you bring the array down to unassign the faulty drive, and a similar time when you bring the new drive into use after the preclear.  The array will remain operable the rest of the time.

[squarefrog]

Another option is to stop the array; unassign the 250GB drive; and start the array with that drive missing.  unRAID will now 'emulate' that drive using the remaining drives plus parity.  That will free up a slot to plug in the new 3TB drive for pre-clearing purposes.  Once the preclear has finished and the drive checks out OK you can assign it in place of the missing drive and unRAID will rebuild the 250GB drive contents onto the new drive.

 

I probably should have mentioned that the 250GB drive doesn't contain any files, as I moved them once I started getting SMART errors. I presume in this case, this second option will allow me to pre-clear the drive inside my box, and not really suffer any downtime?

 

I do want to preclear, as some of the UK couriers aren't particularly careful when handling hard drives and I want to check its reliability before committing data to it. Yep, my parity is 3TB.

[itimpi]

Another option is to stop the array; unassign the 250GB drive; and start the array with that drive missing.  unRAID will now 'emulate' that drive using the remaining drives plus parity.  That will free up a slot to plug in the new 3TB drive for pre-clearing purposes.  Once the preclear has finished and the drive checks out OK you can assign it in place of the missing drive and unRAID will rebuild the 250GB drive contents onto the new drive.

 

I probably should have mentioned that the 250GB drive doesn't contain any files, as I moved them once I started getting SMART errors. I presume in this case, this second option will allow me to pre-clear the drive inside my box, and not really suffer any downtime?

The second option does not give you any significant downtime regardless of whether the drive has files or not.  The plus side is that when the rebuild occurs on introducing the drive to the array you do not really care if it works or fails.

 

As there is no data on the drive then this might be a good time to consider changing the format to XFS (if you have not done so already.  To do so then when the array is stopped you can click on the drive and change the format.  When the array is next started you are given the option to format the drive.  It does not matter at what stage you change the format - it can be done against the emulated drive or the physical drive.  Just make sure it is before any files are added as the format process creates a new empty file system on the drive.

 

I do want to preclear, as some of the UK couriers aren't particularly careful when handling hard drives and I want to check its reliability before committing data to it. Yep, my parity is 3TB.

I agree it is the right thing to do - just though I should mention that is was not ABSOLUTELY necessary in case you were not aware of that.

[squarefrog]

Excellent, I'll do that then!

 

Luckily I did change all my drives to XFS when I first moved to v6.

 

Just out of interest, would you recommend this path if I was upgrading a mostly full 2TB drive to 3TB? I should probably get another drive to keep around as a pre-cleared spare...

[itimpi]

The process should work regardless of the size of the drive involved.

 

In terms of a spare it is worth pointing out that you can never rebuild a drive onto a smaller drive (although a larger one IS allowed).  Therefore the size of the spare drive might determine which of your drives could be rebuilt after failure.

[InlineRanger]

If you plan on growing your array in the future, another idea that will kill two birds with one stone is to get a cheap PCIe x1 SATA III card. The G7 Microservers have a max. total bandwidth of ~750MB/s across the six SATA ports (i.e. 125MB/s per port when all are occupied).

 

Adding a controller will bump up bandwidth to ~185MB/s per port and free up the internal and eSATA ports. With an UnRAID Basic license you can have six array drives, hot plug an app drive in the internal port and hot plug new drives for pre clearing in the eSATA port. My plan is to use a dock w/ fan to make things simple.

 

[squarefrog]

Oh that's an interesting thought! I had planned on replacing the N40L with a Gen8, but they've suddenly become very expensive (new ones coming perhaps?)

 

I'm going to stick with the N40L for another couple of years until I outgrow my current storage, then I'd like to DIY. Providing I can find a nice case thats not too much larger than the N40L and as quiet.

[John_M]

Expensive? They're virtually giving them away here. The Celeron CPU and 4 GB RAM are perfectly adequate for running unRAID.

[PeterB]

I've been using that StarTech dock for a couple of years and it works fine.

 

I have three 5x3 hot swap bays and 16 SATA ports (two of which are eSATA), also my ThermalTake V5 case includes an eSATA socket as standard.

[squarefrog]

Expensive? They're virtually giving them away here. The Celeron CPU and 4 GB RAM are perfectly adequate for running unRAID.

 

I swear 3 days ago they were out of stock everywhere! I agree the N40L is fast enough for unRAID, but I wanted to run plex media server on it which requires extra grunt. The only downside to switching is I couldn't cram all 6 of my drives in which is a little sad.

[kal]

Ive done something similar recently

 

What I did, was remove the failing disk entirely - copied everything off it, removed it from the array (via 'new config', and a parity rebuild), and then physically remove and dump the disk.

 

Means Ive free'd up a slot I can preclear in, and as the disk Im going to preclear is so much bigger then what was in the slot, Ive the option of replacing one of the other disks with it, for a net gain in HD space - and freeing up yet another slot (for future preclears).

 

 

 

[WeeboTech]

The N40L with thebay bios mod allows access to the eSATA port on the back.

An external bay on top of the n40l like the startech or something with a fan and esata would suffice.

It's also reported that the eSATA port supports port multipliers.

[squarefrog]

The N40L with thebay bios mod allows access to the eSATA port on the back.

An external bay on top of the n40l like the startech or something with a fan and esata would suffice.

It's also reported that the eSATA port supports port multipliers.

 

I use the rear eSATA port to run an internal SSD

 

Got 6 drives successfully stuffed in there.

[John_M]

Yes, it just isn't possible to get more than four 3.5-inch disks in the Gen8, though people have done ingenious things with 2.5-inch disks. My own Gen8 has four 6 TB disks for a total of approximately 18 GB of protected storage. Being the basic model, mine didn't come with an optical drive but I found a Blu-Ray drive that fits, which is actually more difficult than just finding an ultra-slim 9.5mm one because the first one I tried - the Panasonic UJ-272 - was a fraction of a millimetre too wide to slide into the slot, so I changed it for the slightly older model UJ-262 and it fits perfectly. I suppose optical drives have limited use with unRAID but it works nicely with the MakeMKV Docker to rip Blu-Rays and DVDs. The optical drive is connected to the motherboard SATA port, as HP intended. To add a pair of SSDs for an unRAID cache pool I fitted an IOCREST (aka SYBA) SI-PEX40063 card into the single PCIe slot. I chose this particular card because it has two 6 Gb/s SATA ports for internal use and two 6 GB/s eSATA ports for external expansion. I chose the 2-lane PCIe version with the Marvell 9235 chip in preference to the similarly priced PCIe x1 version (the SI-PEX40065), based on the 9215, for its increased bandwidth for supporting four ports. I mounted the two SanDisk SSDs on a Schoondoggy bracket. For future expansion I chose an Icy Dock MB561U3S-4SB, which has four 3.5-inch disk trays, a cooling fan, an internal PSU, and both USB 3.0 (not used) and 6 Gb/s eSATA connections. The internal port multiplier works well with the 9235 and I've used the combination to pre-clear four 6 TB disks simultaneously and quite respectable speeds. I'd say that with three simultaneous pre-clears things progressed as fast as directly connected disks. Adding the fourth to the shared eSATA-III link slowed things down a little but performance was more than acceptable. When the external box is full I'll consider adding another to the second eSATA port. I suppose that for the purist external SAS would have been the preferred expansion method, but I'm more than happy with the results I got from 6 Gb/s eSATA and port multiplier and the external box plus controller card were a lot cheaper than SAS-based expansion. Ironically though, for the cost of the external box and the card I could have bought another Gen8 Microserver!

[WeeboTech]

Excellent update John_M.  Anyone know if the SI-PEX40063 is useable/detected under ESX 5.x or 6.x?

[squarefrog]

Hey thanks for the comprehensive reply John_M. Good to know what my future upgrade path can be.

 

The drive I ordered arrived today, so popped it in and it's now pre-clearing. I'm going to be very nervous for the next day or two until I can get the disk in my array and be protected again!

[John_M]

I'm afraid the whole virtualisation thing is unexplored territory for me. I realise it's possible to go in two completely different directions - either running unRAID under ESX or running other OSes under unRAID/KVM - but so far it's just been bare metal for me. Those entry level Gen8 Microservers are an absolute steal at the price and the Celeron processor is more than enough for unRAID, a few plugins and a couple of Dockers but I think I'd need one of the Xeon powered ones to get the best out of virtualisation so it's quite a different proposition, though interesting none the less

[WeeboTech]

I'm afraid the whole virtualisation thing is unexplored territory for me. I realise it's possible to go in two completely different directions - either running unRAID under ESX or running other OSes under unRAID/KVM - but so far it's just been bare metal for me. Those entry level Gen8 Microservers are an absolute steal at the price and the Celeron processor is more than enough for unRAID, a few plugins and a couple of Dockers but I think I'd need one of the Xeon powered ones to get the best out of virtualisation so it's quite a different proposition, though interesting none the less

 

The CPU's can be upgraded. There are links all over the net about people upgrading them.  If you have not needed it by now, you probably do not need it on a Microserver.  I run uTorrent under XP on 3g of ram, unRAID on 4G of ram, slackware development and CentoOS for a gatekeeper with development as other VM's with ESX/Xeon-1220L and 16GB of ram.

I have the startech asmedia esata/sata cards in the PCIe slot. It works under esx, yet if I could get confirmation about the SI-PEX40063 I would switch to those immediately.  2 SSD's internally and 2 eSATA ports, color me interested!

[squarefrog]

OK, i'm about 99% into Zeroing the new 3TB drive. However, just idly reading through posts here trying to learn more about SMART errors, I see this post about replacing Seagate ST300DM001 drives immediately as a precaution for them being very error prone.

 

I check my current array to see if I have any drives, and of course I do... my parity drive 

 

So now I'm worried. I figure I have two options now:

 

[*]Replace the failed 250GB drive, with the soon to be precleared 3TB Red, order a new larger 4TB drive for a new parity drive, then hope nothing bad happens while I wait

[*]Remove the failed 250GB drive from the array, and replace the suspect parity with the 3TB Red. Order a larger 4TB drive, preclear in the now empty slot, then swap the parity 3TB Red with the 4TB. Format the Red and add it to the array

 

I currently have about 3 TB free on my array, so I'm not desperate for the storage right now. I can wait for prices to come down for 4TB. Because of that I think it'd definitely be wise to do the former of the two options.

 

Is there anything in particular I need to do to remove the missing 250GB drive and replace the old parity with the new Red drive?

[WeeboTech]

If the 3tb has been working fine all this time, check the smart values to determine the urgency on it's replacement.

 

I've had a number of the 3TB drives. 1 failed and showed it via smart. The others are almost 2 years old without issue.

You can turn off the spin down timers and issue a smart short and smart long test on the suspect drive after replacing the 250GB or moving the data.

 

I have 3 so I would give it a 33% failure rate so far. Pretty much like the backblaze stats show.

https://www.backblaze.com/blog/3tb-hard-drive-failure/

 

For peace of mind I would start shopping for a replacement after determining urgency on replacing the ST300DM001.

 

 

If the data on the 250 gb drive is not important or has been moved off to the other drives, I might elect to replace parity earlier, but probably would not.

Depends on age of 3TB drive and if reallocated, uncorrectable or pending sectors are showing up. (but that is me).

 

I would probably order and preclear a drive for parity that I can live with for a while.

For me, these days I go to the 6tb hgst drives. They get 225MB/s on the outer tracks. YMMV.

 

 

If the goal is cheapest price per GB available and you don't need the expanded space right now, then put the new drive in service as parity while shopping for an additional drive.

 

 

As far as removing the missing 250GB drive and replacing old parity with the new drive, I would move data off the 250gb drive with rsync to any spare space on the other drives.  Capture a printout or image of the current drive layout, then rebuild the array via new config from scratch utilizing the new drive as parity.

[WeeboTech]

FWIW, and I realize no one is all that rich, the Dell E4300,E6400,6410,6510 have an eSATA port and can be purchased used on eBay cheaply. They make good preclear workstation machines.  Using one of external bays or whatever you have available with the eSATA port you can use it as a standalone preclear work station.

 

For almost an equivalent amount you might be able to score a used N40L, throw in an ICYDOCK Duoswap on top and use that as a backup for the most critical files along with a preclear station. This is what I did with my older N54L's

 

http://www.icydock.com/goods.php?id=141

[neilt0]

I run unRAID on an N54L and it's fast enough for my purposes.  I have 10 drives attached (3 are 2.5" and one is external).

I use an Adaptec 1430SA.

 

I also have a Gen8, but am not running unRAID on it yet.  I think I paid £180 with £100 cashback.

 

Also used to have an N36L, but sold it recently. 

[squarefrog]

If the 3tb has been working fine all this time, check the smart values to determine the urgency on it's replacement.

 

Yes, I can probably do this. I'm not entirely sure what I should be looking for when reading the smart report, is there a sticky post or wiki somewhere that explains what to look for?

 

 

For peace of mind I would start shopping for a replacement after determining urgency on replacing the ST300DM001.

 

I had planned on replacing the ST300DM001 with a larger better quality further down the line anyway so this is no real hardship.

 

If the data on the 250 gb drive is not important or has been moved off to the other drives, I might elect to replace parity earlier, but probably would not.

Depends on age of 3TB drive and if reallocated, uncorrectable or pending sectors are showing up. (but that is me).

 

I had already moved all the data off the 250GB, so it's been sat unused for a few months, until a reasonable deal came along.

 

I would probably order and preclear a drive for parity that I can live with for a while.

For me, these days I go to the 6tb hgst drives. They get 225MB/s on the outer tracks. YMMV.

 

If the goal is cheapest price per GB available and you don't need the expanded space right now, then put the new drive in service as parity while shopping for an additional drive.

 

As I'm still only about 60% utilised, I can't quite justify the leap to 6TB, as much as I'd like to! I think it makes sense to just reallocate it as a parity and wait for prices to drop further, or for me to outgrow my current storage.

 

As far as removing the missing 250GB drive and replacing old parity with the new drive, I would move data off the 250gb drive with rsync to any spare space on the other drives.  Capture a printout or image of the current drive layout, then rebuild the array via new config from scratch utilizing the new drive as parity.

 

Am I right in thinking this is achieved by stopping the array, and going to Tools > New Config? The alert was a bit alarming - so just to confirm it wont format my drives, but simply trigger the rebuild of the parity? Any harm in removing the missing 250GB and swapping the parity in one step?

[WeeboTech]

If the 3tb has been working fine all this time, check the smart values to determine the urgency on it's replacement.

 

Yes, I can probably do this. I'm not entirely sure what I should be looking for when reading the smart report, is there a sticky post or wiki somewhere that explains what to look for?

 

There are plenty of hints around the board. What comes to mind are any pending sectors (which are a timebomb) significant reallocated sectors or an increasing amount, any uncorrectable sectors. High load cycle count.  From what I read over 600,000 is cause for concern.

 

 

When My drive started to fail, there were pending sectors. I quickly replaced it then did some tests.

I did a smart long scan and badblocks scan. the pending sectors went away. I put it back in service as parity.

A short time later the pending sectors were growing and I took it out of service for a RMA.

 

The smart long scan will read every sector.

do a smartctl to capture the logs first, then a smart long scan. (turn off the spin down timer first), then capture a smartctl log for comparison.

 

There's some information here, but not enough.

http://lime-technology.com/wiki/index.php/Understanding_SMART_Reports

 

I would suggest scanning the forums for other conversations.

If you are running unRAID 6, there should be some intelligence to warn you if there are issues.

 

I would probably order and preclear a drive for parity that I can live with for a while.

For me, these days I go to the 6tb hgst drives. They get 225MB/s on the outer tracks. YMMV.

 

If the goal is cheapest price per GB available and you don't need the expanded space right now, then put the new drive in service as parity while shopping for an additional drive.

 

As I'm still only about 60% utilised, I can't quite justify the leap to 6TB, as much as I'd like to! I think it makes sense to just reallocate it as a parity and wait for prices to drop further, or for me to outgrow my current storage.

 

Considering your usage pattern, it doesn't make sense to get another drive yet.

It makes sense for your peace of mind, to replace the parity drive with the newly precleared drive.

 

As far as removing the missing 250GB drive and replacing old parity with the new drive, I would move data off the 250gb drive with rsync to any spare space on the other drives.  Capture a printout or image of the current drive layout, then rebuild the array via new config from scratch utilizing the new drive as parity.

 

Am I right in thinking this is achieved by stopping the array, and going to Tools > New Config? The alert was a bit alarming - so just to confirm it wont format my drives, but simply trigger the rebuild of the parity? Any harm in removing the missing 250GB and swapping the parity in one step?

 

Make note of where drives are attached. Capture the screen if you can.

Do the new config.

Assign all DATA drives first. Do not assign parity yet.

Start array. Make sure all drives are mounted.

Once you have verified all data drives are where they should be.

You can stop the array.

Assign the new parity drive

Start the array.

Let the parity sync occur.

 

Many people make the mistake of accidentally assigning a data drive to the parity slot thus loosing that data drive.

Therefore I suggest you check all data slots first before ever assigning a parity drive.

 

 

I would not remove the 250 gb and assign the data and parity drives in one setting.

I usually do it piece meal. 

New config, assign only the drives I want to keep.

start array.

verify.

Nothing should require formatting at this point in time since nothing is new.

when good, stop array

add in parity.

start array.