Ready for some HD upgrades after 10 years running UnRaid


Recommended Posts

Hi Everyone,

I had a small issue a month ago which got resolved very quickly, thanks to Jon and Tom, awesome Guy's Thanks.

 

After not really touching my 2x 40Tb Towers for nearly 10 years after setting them up, other than loading/deleting Movies and TV Shows,

I use Plex and Dune media player to Stream to all my devices.

Blowing out the heatsinks and dust bunnies...every month, have a reminder set in my phone so I do not miss this very important thing.

Probably why I have not had too many disk failures..

 

I have 20x 2Tb drives, plus Parity in each enclosure which I customized myself.

Details of my Tower1 below.

M/B: ASUSTeK Computer INC. - P5K-E

CPU: Intel® Core™2 Duo CPU E7500 @ 2.93GHz
HVM: Enabled
IOMMU: Disabled
Cache: 64 kB, 3072 kB
Memory: 4 GB (max. installable capacity 4 GB)*
 
Tower2 details,
System: ASUSTeK Computer INC. - P5Q DELUXE
CPU: Intel(R) Core(TM)2 Duo CPU E7500 @ 2.93GHz - 2.933 GHz
Cache: 64 kB, 3072 kB
Memory: 2048 Module (max. 4 GB)
 

 

Only had to replace 4x 2Tb drives in all that time, WD and Samsung were originally installed, 1x WD and 3x Samsung failed. Not bad overall... Replaced with WD..

 

Now, it is time to update both Towers with Cache drives and some bigger drives, maybe some RAM for Tower2??.

About 12 of the drives over the 2x Towers are up to over 5 years Power On in the Smart details. No errors yet, but it is time to start planning.

Got a little cash stashed, so let's get these Babies all freshened up.

 

So far looking at a 256Gb WD Green SSD for the Cache, single one only, unless someone can convince me Dual Cache is better?

Was going to start with 6Tb Seagate Ironwolf as they are a little cheaper than the WD Red, by 30 Swiss Francs, But I am kind of drawn to the WD as they have been so good for me so far..

I went through the Forum and did not find any issues with either hard drive.

 

So the plan would be to replace Parity in each Tower to 6Tb first, then move onto the data drives one by one.

So maybe I use the WD Red for Parity, and the Seagate for data??

 

As I just upgraded from v5.0.5 to v6.3.5, the drives are in reiserfs, should I look at going to XFS??

 

Anything else I should look at doing??

 

Appreciate your advice and comments.

 

Much appreciated.

Steve

Link to comment

Hi, that's an amazingly long and good run - hope it continues!

 

I would look at moving from ReiserFS to XFS, especially if your drives are near capacity.  In practical terms of server usage I don't think the 7200rpm Seagates will seem all that much faster than the WD-Reds - writes will continue to be slowed down by the 2TB drives for now and either a Seagate or a WD Red are capable of saturating a 1GB network on reads.   I'd probably go with whatever is cheapest - but personally I'd be looking at 8TB drives as they are the cheapest per TB here in the US and there are some tremendous deals if you are willing to shuck drives from USB enclosures.  I've had good luck with WD Reds for what it's worth, though.

 

I'd try to get Tower2 up to 4GB as unRAID 6 uses more memory.  Used purely as a NAS you've probably got a few more years on those motherboards and CPUs - but they're getting old as well and you should start thinking about the need to replace them eventually too.  Is there a reason you feel like you need cache drives?  Writes to the new, higher capacity drives will be a lot faster than the 2TB drives.

Link to comment
1 hour ago, HK-Steve said:

 

So far looking at a 256Gb WD Green SSD for the Cache, single one only, unless someone can convince me Dual Cache is better?

 

 

I don't know about the single vs multiple cache drives, but I think you want at least 500GB or more for cache drive.  My appdata is 186GB with only half a dozen of dockers.

Edited by themaxxz
Link to comment
1 hour ago, tdallen said:

Hi, that's an amazingly long and good run - hope it continues!

 

I would look at moving from ReiserFS to XFS, especially if your drives are near capacity.  In practical terms of server usage I don't think the 7200rpm Seagates will seem all that much faster than the WD-Reds - writes will continue to be slowed down by the 2TB drives for now and either a Seagate or a WD Red are capable of saturating a 1GB network on reads.   I'd probably go with whatever is cheapest - but personally I'd be looking at 8TB drives as they are the cheapest per TB here in the US and there are some tremendous deals if you are willing to shuck drives from USB enclosures.  I've had good luck with WD Reds for what it's worth, though.

 

I'd try to get Tower2 up to 4GB as unRAID 6 uses more memory.  Used purely as a NAS you've probably got a few more years on those motherboards and CPUs - but they're getting old as well and you should start thinking about the need to replace them eventually too.  Is there a reason you feel like you need cache drives?  Writes to the new, higher capacity drives will be a lot faster than the 2TB drives.

Thanks tdallen, I also hope the good long runs continue.

I do understand that the older drives will slow the speed, but will try to upgrade Tower1 with new drives, copy all the data to Tower1 then convert the older Tower2 to XFS,

then can upgrade the old Tower2 to newer drives also, will take some time. Budget is good, but 20x replacement drives is some serious coin..

That is why I was looking at the 6Tb drives, but will look around here and see what I can find before I jump.

 

Will do for Tower2 RAM, an easy upgrade..

 

I was just reading about the Cache drives and how they speed up data transfer, So more curiosity rather than a need.

 

Much appreciate your advice and comments.

 

Stay tuned.

 

Cheers

Steve

Link to comment
52 minutes ago, themaxxz said:

 

I don't know about the single vs multiple cache drives, but I think you want at least 500GB or more for cache drive.  My appdata is 186GB with only half a dozen of dockers.

I do not use any Dockers or VM's.. all disabled at the moment.

Is purely used as a NAS...

But 500Gb is also good if I need the Cache drive to speed up the transfer of data.

 

Thanks

Steve

Link to comment

The other drive brand to consider is HGST. Although owned by WD, they operate pretty independently. HGSTs have gotten best results in long term studies. Their NAS drives are 7200RPM and I've had excellent results with them. Much better than Seagate IMO.

 

But the best deal IMO is the best buy WD easystore 8T drives. They contain WD Reds. Prices vary between $170 and $200. Shucking is pretty easy without messing up the external enclosure (so can be re-assembled if needed if the drive is defective.) 

 

I use my cache drive for VMs and Dockers only. I prefer to have the writes be protected right away, not have to wait until overnight. And if I am copying a lot of data, I can turn on turbo write and get very fast writes directly to the array.

 

But I'll bet if you start playing with them, you'd find the Docker apps and VMs very useful! I'd get the cache for that! I've replaced my desktop with a VM with video passthrough. One of the big advantages IMHO of the extremely fast server accesses. No network. I can copy files at over 200 MB/sec to unassigned devices. And with turbo write, might get close to that to the array.

  • Upvote 1
Link to comment

There are pros and cons to using a cache drive to cache writes to the array.  In the days when writes to the parity protected array were 10MB/s and writes to the cache drive were 50MB/s and servers had 1GB of memory, the cache drive was a pretty big deal and the fact that your data wasn't protected by parity seemed like a fair trade-off.  Now that writes to the array are commonly > 50MB/s with today's large, fast hard drives and many smaller writes are cached in system RAM at full line speeds it isn't as clear.  Also, with Turbo Write you can now achieve very high write throughput directly to the array if you need.  Further, how often do you sit and watch writes to your server?  It probably doesn't matter how fast they are if no one is watching...  Write caching is right for some people and some use cases, but it isn't universally a better thing than writing directly to the array.

 

I no longer cache writes to my array, I use my cache drive strictly for Dockers and playing with VMs.

 

If you buy 8TB drives, you only need 6 of them to replace that entire 20 drive array of 2TB drives... 

  • Upvote 1
Link to comment

I would recommend the new WD-Red 10TB Pro because they are more energy-efficient then the 6TB drives, they have 5 years warranty and running also at 7200U/min.

Or you use the normal WD-Red 10TB with 3 years warranty and 5400U/min.

I dont recommend Seagate because of a bad experience over the last 17 years - but thats only my opinion ^_^

 

For the Cache-SSD its important to use a model with a high TBW value because its used for Dockers, VMs, Transcoding from Plex and transfering Data to the array.

A normal Desktop-SSD like the Samsung EVO 850 has a TBW of approx. 75TB (256GB-model).

Because the Cache-drive writes a lot of data, the TBW is important.

My Transcent SSD has a TBW of 280TB and has written approx 46TB over the last 2 years.

I would buy a Kingston SSDnow KC400. The 256GB model has a TBW of 300TB, the 512GB model has a TBW of 800TB, both with 5 years warranty.

 

Edited by Zonediver
Link to comment

The problem with the 10tb drives is cost. The cheapest I saw was the HGST at $344, while the WD 10T pros are $450. The 8T easystore (WD Red 8T) is $180. So you are spending double or more and only getting 25% more space. If you are buying 1 or 2, and need the extra performance and density, you might justify. But for large media servers needing a half dozen or more, where rotational speed means almost nothing, the cheaper ones make more sense.

 

In every age there is a drive in the sweet spot of $/TB. Right now that is clearly 8T.

  • Upvote 1
Link to comment

Thanks Guy's, appreciate your advice.

Not going to worry about the Cache for now, will get this upgraded, then look into Dockers and VM later.

 

I can not get the WD easystore 8T drives here, looked everywhere today..Bummer.

 

LaCie Porsche Design Desktop Drive Design P'9233  (8000GB) I found for a good price, but I have no idea what drive is inside,

from the specs sounds like a 5400rpm drive.

Could find nothing else, the 1Tb enclosure I have myself has a Samsung in it, some of the video's on YT show a Seagate in the 4Tb units. So no idea for the 8Tb.

 

Much appreciated.

 

 

 

Link to comment
On 11-10-2017 at 7:25 AM, Zonediver said:

 

A normal Desktop-SSD like the Samsung EVO 850 has a TBW of approx. 75TB.

 

 

Thanks for the good info.

I knew about SSD write limitations, but never really thought about it.  My 14month old 500GB EVO 850 has done 21TB, and it seems to be spec'd for 150TBW.  ( http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/850evo.html )

 

Link to comment
1 hour ago, themaxxz said:

 

Thanks for the good info.

I knew about SSD write limitations, but never really thought about it.  My 14month old 500GB EVO 850 has done 21TB, and it seems to be spec'd for 150TBW.  ( http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/850evo.html )

 

 

Yep - thats the point. Samsung SSDs are good and fast but the TBW is low. Even the 500GB model has only 150TB TWB. For a cache-drive not really much.

But normaly, the "real" duration "should be" at least 1,5x-2x of the specified TBW. Some SSDs have even more the 4x of the specified TBW - but that's only a theory because its not guranteed ^_^

Edited by Zonediver
Link to comment
22 hours ago, SSD said:

The problem with the 10tb drives is cost. The cheapest I saw was the HGST at $344, while the WD 10T pros are $450. The 8T easystore (WD Red 8T) is $180. So you are spending double or more and only getting 25% more space. If you are buying 1 or 2, and need the extra performance and density, you might justify. But for large media servers needing a half dozen or more, where rotational speed means almost nothing, the cheaper ones make more sense.

 

In every age there is a drive in the sweet spot of $/TB. Right now that is clearly 8T.

 

Here in Austria, the price is very high:

8TB WD-RED: € 276.- ($ 326.-)

8TB WD-RED Pro: € 322.- ($ 381.-)

10TB WD-RED: €375.- ($ 444.-)

10TB WD-RED Pro: € 421.- ($ 498.-)

...thats not fun >:(

The cheapest at the moment is the WD-RED 3TB with only € 95.- ($ 112.-) but that model is useless (for me) ^_^

 

Next point: Energy efficiency:

8TB WD-RED Pro: 7,2W (operation) / 5,1W (idle)

8TB WD-RED: 6,4W (operation) / 5,1W (idle)

10TB WD-RED: 6,2W (opertion) / 2,8W (idle)

10TB WD-RED Pro: 5,7W (operation) / 2,8W (idle) - 4TB WD-RED: 4,5W (operation) / 3,3W (idle)

So the 8TB model is not an option compared with the 10TB model...

 

Edited by Zonediver
Link to comment

I easily forget that we have an international set of users here. My apologies. And based on the data provided, the 10TB units seem only incrementally more expensive in Austria. In the US, it is the 8T external drives that can be shucked that provide the low prices. Is something similar avail in Austria?

 

8T Red (in external case) &179 USD

8T Red $274 USD

8T Red Pro $330 USD

10T Red $380 USD

10T Red Pro $450 USD

 

If you ignore the external one, the cost of the 10T drives is more per T, but not unreasonably so. But if you include the external 8T, it is hard to recommend anything else in the US. 

 

The HGSTs are my choice for 7200 RPM

HGST 8T NAS - $270 USD

HGST 10T NAS - $344 USD

 

Cheaper and better in my experience. Not sure if those are an option in other countries.

 

I don't tend to focus as much on the energy usage. Disks are very often spun down.

Link to comment

Hi Guy's, was doing a parity check today and got an error. 

2017-10-12, 11:34:07 3 hr, 23 min, 56 sec Unavailable Canceled

257 errors

 

Now my hard drieve 16 is offline, Say's Device is offline, content emulated... with a red X instead of green dot before Parity check started.

 

What is the next step for me, as I was going to install new upgraded hard drives tomorrow. 8Tb's.

 

Thanks

Steve

Link to comment

@HK-Steve 

 

(This is in regards to the drive failing)

 

If you've been inside the server and disturbing cables, it may not have failed, but simply lost connection. If that's what's happened you could rebuild unto itself.

 

If not and the drive failed because it's getting old, and there are a bunch of similar disks inside, you have what I'll call a delicate array. It contains a bunch of older disks and now one had failed, so the risk of another failure is higher than normal.

 

There are a couple options. One is to replace the disk with one of equal size (or larger but still no bigger than parity). This is straightforward, but I would be a bit concerned for another failure. If you have full backups, could be a good option. But not what I'd do.

 

Second, you could do what's called a parity swap disable. In this, you are upsizing parity to an 8T/10T drive and then using the old parity to rebuild the failed disk onto. Similar concerns with above. But does upsize parity in the process. See below for more info on this option.

 

Third option, and what I would do, is to add a new 8T/10T to the server, partition and format it to unRaid spec (unassigned devices cannot partition it properly, but after it is partitioned can format and mount it for you). You can then copy your data from the emulated disk to the new 8T/10T disk. You can copy your most critical data first. Another disk can still potentially drop offline, but whatever has already been copied is safe. If doing a rebuild operation, until the entire disk is copied over, you have no idea what files might be restored, partially restored. It would be a huge mess to sort out of a drive failed in the middle. If you do this, you could follow a similar process with all your functioning disks, one at a time, copying the contents to new 8T/10T drives. (It will be fast as parity is not being updated). Once all data is copied over to new disks, and you complete whatever QA process, you can remove the old disks, install all the new ones, and build parity. Again, this is what I would do. Few more steps for you, but cuts the upgrade by about half. If you have enough room to add a couple new drives at once, you can copy data to them in parallel and save even more time. And this would make it easier to combine the contents of 4T drives onto the larger new drives.

 

Another option would be to copy the emulated data to an external disk attached to a workstation. At least you'd have the data backed up. You could then do the parity swap disable to upsize your parity and make your array whole again. And then incrementally rebuild first disk. And then copy data from the second disk to fill it up to desired level. And then rebuild the third disk. Etc. Little slow but would work.

 

What disks to buy is up to you, but I'd pull the trigger quickly to get your array protected again.

 

Good luck.

Link to comment

Thanks SSD,

I have not moved or disturbed anything in the Tower1, I had just finished copying over about 2Tb of movies over 3 hard drives, disks 2, 3 and 4.

Then wanted to do Parity Check ready for tomorrow Parity disk change over to 8Tb.

 

Agree with the Fragile disks, this is why I am trying to move forward and get this updated quickly.

 

I am nearly in the frame of mind to build a new Tower3 tomorrow with New motherboard, CPU and Ram, case and install 6x 8TB drives, and replace the Tower1 completely.

Then I can also complete the XFS conversion also which needs to be done on Tower1 anyway..

Copy everything off Tower1 to Tower3.

Then get new drives for the failing Tower1 and start fresh when funds allow..

 

Link to comment

It looks like a pretty sweet setup! The MB is an EEB form factor, and you just need to make sure your case will handle. Might run pretty warm, and dim the lights a bit.

 

You might peek at the new Coffee Lakes (see link below). Hex core. Pretty well priced as trying to combat AMD Ryzen / TR. I'd look at one of the i7s or the i5-8600K. Might be similar price to the rig you are looking at, even with memory added. (no need to buy new heat sinks and would be much more power efficient). No ECC memory though. You'd have a lot fewer cores, but each one would be a lot faster.

 

https://arstechnica.com/gadgets/2017/10/intel-coffee-lake-8700k-review/

 

Copying from tower1 to tower3 is a good idea. If you can copy data off of each disk (read each disk once), that would be my preference. Performing multiple whole disk operations (like parity checks and rebuilds) before doing the copying would just add more risk of failure.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.