Afternoon All,

 

I'm seriously considering coming back to unRaid from Flexraid.  I was a user of unRaid from 4.7 up to 5.5, but then decided to try something else.  That being said, I have some questions regarding unRaid 6.

 

1.  Has performance of the array (not counting cache drives) increased?

2.  How does the cache pool work?  eg - If I have a 240gb, 256gb, and 480gb ssd in the pool how much space do I have in the pool?

3.  How does the virtual networking compare to VMWares?  My system currently has 6 NIC's in it.

4.  Is dual parity on the roadmap?  If not, will unRaid see the software raid on my supermicro board (I assume not)

5.  I bought my license (I hope my usb still works!) back in 2011 when 4.7 was current... can I simply upgrade to the latest version?

 

I currently run esxi on my server, and while I realize that unRaid runs fine under vmware, I'm looking to simply my life.  I no longer run multiple VM's, instead just running a download VM (NZBGet/Deluge) and a Python Server (CP, HP, Mylar, Sonarr, etc)... so I feel that these can easily be integrated into unRaid's VM implementation.

 

Lastly, what kind of resources does unRaid require now that it has these additional capabilities?

 

Thanks in advance!  I'm sure I'll have more questions as I plan this migration out

 

~Spritz

Afternoon All,

 

I'm seriously considering coming back to unRaid from Flexraid.  I was a user of unRaid from 4.7 up to 5.5, but then decided to try something else.  That being said, I have some questions regarding unRaid 6.

 

1.  Has performance of the array (not counting cache drives) increased?

2.  How does the cache pool work?  eg - If I have a 240gb, 256gb, and 480gb ssd in the pool how much space do I have in the pool?

3.  How does the virtual networking compare to VMWares?  My system currently has 6 NIC's in it.

4.  Is dual parity on the roadmap?  If not, will unRaid see the software raid on my supermicro board (I assume not)

5.  I bought my license (I hope my usb still works!) back in 2011 when 4.7 was current... can I simply upgrade to the latest version?

 

I currently run esxi on my server, and while I realize that unRaid runs fine under vmware, I'm looking to simply my life.  I no longer run multiple VM's, instead just running a download VM (NZBGet/Deluge) and a Python Server (CP, HP, Mylar, Sonarr, etc)... so I feel that these can easily be integrated into unRaid's VM implementation.

 

Lastly, what kind of resources does unRaid require now that it has these additional capabilities?

 

Thanks in advance!  I'm sure I'll have more questions as I plan this migration out

 

~Spritz

 

Can't answer all of your questions, but I can answer a few.

 

2. The cache pool uses BTRFS raid so while I don't have the exact answer there might be some tool out there to point you in the right direction.

4. Dual Parity is on the road map and in theory coming in 6.2 (Which might be this year or not...)

5. Yes you should be good to just follow the instructions to update from 5.5 to 6.1.7

 

 

2.  How does the cache pool work?  eg - If I have a 240gb, 256gb, and 480gb ssd in the pool how much space do I have in the pool?

2. The cache pool uses BTRFS raid so while I don't have the exact answer there might be some tool out there to point you in the right direction.

btrfs disk usage calculator

Awesome, thanks for the replies.  To be honest, the community is a large part of the reason that I'm looking to come back.

 

Very helpful and knowledgeable.

 

~Spritz

1.  Has performance of the array (not counting cache drives) increased?

 

While Unraid performance has not increased in a while (I believe somewhere during v4.4 or v4.5 there was a big change), and mostly due to higher density platter disks, it's much better than a few years ago, It's not difficult to get sustained writes of 50 to 70 MB/s during at least the first half of modern disks, combine that with turbo write mode and using more RAM as write cache you can have very good performance for most situations without using a cache drive.

 

While Unraid performance has not increased in a while (I believe somewhere during v4.4 or v4.5 there was a big change), and mostly due to higher density platter disks, it's much better than a few years ago, It's not difficult to get sustained writes of 50 to 70 MB/s during at least the first half of modern disks, combine that with turbo write mode and using more RAM as write cache you can have very good performance for most situations without using a cache drive.

 

I'm intrigued by this turbo write mode and write cache?  Can you elaborate?

 

~Spritz

I believe your Download VM and Python VM can be run in Dockers, so no VM's at all!

I believe your Download VM and Python VM can be run in Dockers, so no VM's at all!

 

That would be the plan, though I still need to do some more reading on how Dockers work, and how to build them.  I have a bit of an odd setup in that I run 3 instances of Couchpotato (Our Movies, Chick Flicks, and Kids movies), and I use Sonarr as my TV downloader.

 

It'll be interesting to see what they can do.  Now the real issue will be the migration, and how to minimize downtime.

 

~Spritz

3 instances of couch potato would be easy with docker.

I'm intrigued by this turbo write mode and write cache?  Can you elaborate?

 

~Spritz

 

Turbo Write can be very useful for certain situations, like large writes, I have it always on for my archive servers, servers that are always off and only turned on once a week to move some data, so I don’t care about disk spindown and can copy data at or very close to gigabit speed. Note that for turbo write to work well the server can’t have controller/interface bottlenecks, a good reference point is your parity check speed, as write speed with turbo on can’t be higher than that.

 

As for using RAM for write cache, since V6 is 64bits you can use all your free RAM to help with short writes, by default Unraid uses 20%, on my always on daily server I have 16GB and changed it to 99%, so and since my daily writes are usually between 10 and 15GB, all are done at gigabit speed.

 

Between turbo write and RAM cache I’d say 99% of all writes to my servers are at or very close to gigabit speed, 110MB/s.

 

3 instances of couch potato would be easy with docker.

 

That's great to hear!  One last question (for now), is once a docker is installed, am I dependent on the docker creator to update the software or can I just use the softwares built in auto-update.

 

Thanks again!

 

~Spritz

Well it depends on which creator's version you use.  Our containers at linuxserver.io autoupdate on restart.  Others are dependent on the author triggering the build

Well it depends on which creator's version you use.  Our containers at linuxserver.io autoupdate on restart.  Others are dependent on the author triggering the build

 

Linuxserver.io looks awesome, you seem to hit all the big projects!  The only ones I see missing are Calibre Server, Maximus (https://github.com/mescon/Muximux), and the Google Play uploaders (photo and music) but I can either learn how to build a docker or throw up a VM.

 

Another question is the best way to configure my "extra" drives.  I'll explain a bit better...

 

Currently I have my storage array setup with 2 parity drives and 17 data drives.  The migration of that to unRaid is straight forward (relatively speaking).  However, I also have a 240gb, 256gb and 512gb ssd as well as a WD 750gb Black drive that I use outside of the storage array for various tasks.  Currently both the smaller SSD's are used for my VM datastores, and the 512gb SSD is used for the Flexraid cache drives.  Lastly the 750gb Black is used as my scratch drive for my downloads.

 

*phew* all that to say, is I'm not sure how that would translate into the new unRaid setup... suggestions?

 

Thanks again for all your help.

 

~Spritz

Set the 512gb SSD as the cache drive. Download the unassigned devices plugin and mount the 240 gb as vmdisk-1, mount the 256 gb as vmdisk-2 and finally mount the 750 fab disk as "downloads". Of course you can change the mount point names to anything but that is exactly how I have my server setup.

Set the 512gb SSD as the cache drive. Download the unassigned devices plugin and mount the 240 gb as vmdisk-1, mount the 256 gb as vmdisk-2 and finally mount the 750 fab disk as "downloads". Of course you can change the mount point names to anything but that is exactly how I have my server setup.

 

Only problem with that is if the fab disk is mounted after the docker service starts then you won't be able to use it in containers unless you restart the docker service.

 

One way around this would be to put the docker.img on the fab disk and then the docker service can't start until the disk is mounted, solving the problem.

Since you can get away w/o actually using a VM, i'd just use the 512GB SSD as a cache drive and set up everything to run off that.

 

Keep it simple where you can, no need to add all these (semi)manually mounted devices.

Set the 512gb SSD as the cache drive. Download the unassigned devices plugin and mount the 240 gb as vmdisk-1, mount the 256 gb as vmdisk-2 and finally mount the 750 fab disk as "downloads". Of course you can change the mount point names to anything but that is exactly how I have my server setup.

 

Only problem with that is if the fab disk is mounted after the docker service starts then you won't be able to use it in containers unless you restart the docker service.

 

One way around this would be to put the docker.img on the fab disk and then the docker service can't start until the disk is mounted, solving the problem.

 

Interesting, i guess it's a good thing my docker.img is on the UD disk as well.

Set the 512gb SSD as the cache drive. Download the unassigned devices plugin and mount the 240 gb as vmdisk-1, mount the 256 gb as vmdisk-2 and finally mount the 750 fab disk as "downloads". Of course you can change the mount point names to anything but that is exactly how I have my server setup.

 

That was sort of my thinking... I may pick up another cheap 500ish gig SSD so that I can have a cache pool... seeing as they've plummeted in price.

 

Only problem with that is if the fab disk is mounted after the docker service starts then you won't be able to use it in containers unless you restart the docker service.

 

One way around this would be to put the docker.img on the fab disk and then the docker service can't start until the disk is mounted, solving the problem.

 

I assume that once the docker image loads, it's running primarily in memory, thereby negating the fact that it would be running off of a HDD.

 

Since you can get away w/o actually using a VM, i'd just use the 512GB SSD as a cache drive and set up everything to run off that.

 

Keep it simple where you can, no need to add all these (semi)manually mounted devices.

 

I agree with you in theory (simplification is a driving force behind this migration), but I'm afraid having everything running off the cache drive will overwhelm the cache drive.  It's why in the past I've had a download scratch disk.  Though I imagine that if I run a docker with nzbget it should have an option to point to a download location... which then could be another drive (or am I out to lunch?).

 

Thanks for all the comments

 

~Spritz

SSD's excel at high I/O applications, I'm guessing you'll be just fine. I run my cache disk off an old laptop drive and have no performance issues with several dockers (one being plex).

I assume that once the docker image loads, it's running primarily in memory, thereby negating the fact that it would be running off of a HDD.

 

My docker.img is 30GB so I hope it isn't resident in memory.    The docker.img is like a virtual btrfs drive.  I run mine off a drive that is mounted outside the array.

Thanks again for all the help.

 

Can anyone comment on how the virtual networking done in comparison to esxi?  For example, in esxi, I can create virtual switches and assign physical nics to those switches, and then logically assign my vm's to them.  Is something similar available in the unRaid virtual world?

 

For migration, has anybody else done a large scale migration to unRaid?  I have ~35TB worth of data and want to minimize downtime.  Building another server is not really an option.

 

~Spritz

Thanks again for all the help.

 

Can anyone comment on how the virtual networking done in comparison to esxi?  For example, in esxi, I can create virtual switches and assign physical nics to those switches, and then logically assign my vm's to them.  Is something similar available in the unRaid virtual world?

 

For migration, has anybody else done a large scale migration to unRaid?  I have ~35TB worth of data and want to minimize downtime.  Building another server is not really an option.

 

~Spritz

 

You can make and assign bridges internally, however, best performance is had when doing a pass through of a physical card to a VM (like pfSense, my main usage case). I believe the docker images can all run on a single internal bridge or two depending on how you want to configure unRAID. Might want to create a thread just for that issue to attract LT staff or a more experienced user that has played with that exact thing to chime in on the thread.

Ok, as I continue down this rabbit hole of moving back to unRaid, 2 more questions have popped up.

 

1.  To cache pool or not to cache pool.  I've read mixed experiences with the cache pool, and am unsure of which direction to go.  Will unRaid notify you if the cache pool is running in a degraded state?  That may sway my decision.

 

2.  Does the unassigned devices script auto mount devices (or automount defined devices), or is this an entirely manual process?

 

Thanks again for all the great feedback

 

~Spritz