Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Software - Troubleshooting and Discussion > macOS > The return of ZFS

The return of ZFS (Page 3)
Thread Tools
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
Feb 6, 2012, 12:18 PM
 
Originally Posted by mduell View Post
My big concern here is how it will handle the directory hard links that TM uses; I don't think ZFS supports them.
That's my worry as well. Apparently there is a workaround that basically involves making a big .sparsebundle HFS+ image on any drive. Useful if I have to still support backups from the SSD or the MBA but still have the main volume of data come from the HDD, but not otherwise. For anything like TM to work, it means figuring out snapshots, zfs send/receive and some thinning logic into a script - for a gain that is very intangible at this point. For all that millions of files in one directory as a hard link hack looks scary, Apple has tested this and it has seemingly worked for Apple's customers since 10.5.

Originally Posted by mduell View Post
Native write caching is on a per-drive basis, so it impacts all file systems on the drive. The impact depends on your usage pattern, the biggest impact being on small random writes. It's possible whatever mechanism normally enables/disables write caching will leave it enabled.
My boot drive not having write caching seems like a very bad idea. Since the HDD is also partitioned (a BootCamp partition), I need ZIL/L2ARC for decent performance against the HDD.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
mduell
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status: Offline
Reply With Quote
Feb 6, 2012, 01:10 PM
 
Originally Posted by P View Post
That's my worry as well. Apparently there is a workaround that basically involves making a big .sparsebundle HFS+ image on any drive.
I didn't think of that, but now that you mention it that's what it does for network shares too. Seems OK.

Originally Posted by P View Post
My boot drive not having write caching seems like a very bad idea. Since the HDD is also partitioned (a BootCamp partition), I need ZIL/L2ARC for decent performance against the HDD.
Since it's an SSD I wouldn't be too worried about the impact of write caching off.
     
CharlesS
Posting Junkie
Join Date: Dec 2000
Status: Offline
Reply With Quote
Feb 6, 2012, 06:46 PM
 
Originally Posted by P View Post
That's my worry as well. Apparently there is a workaround that basically involves making a big .sparsebundle HFS+ image on any drive. Useful if I have to still support backups from the SSD or the MBA but still have the main volume of data come from the HDD, but not otherwise. For anything like TM to work, it means figuring out snapshots, zfs send/receive and some thinning logic into a script - for a gain that is very intangible at this point. For all that millions of files in one directory as a hard link hack looks scary, Apple has tested this and it has seemingly worked for Apple's customers since 10.5.
Since snapshots aren't coming until the Gold version of ZEVO anyway, why not just wait for that to come out? They have to know that this would be a popular use case, so maybe they'll provide some sort of TM-alike functionality so that you don't have to write that script.

Ticking sound coming from a .pkg package? Don't let the .bom go off! Inspect it first with Pacifist. Macworld - five mice!
     
BLAZE_MkIV
Professional Poster
Join Date: Feb 2000
Location: Nashua NH, USA
Status: Offline
Reply With Quote
Feb 6, 2012, 07:13 PM
 
Originally Posted by mduell View Post
that's what it does for network shares too. Seems OK.
Just pushed the boulder down the hill. I had my remote time machine backup fail because the disk image got corrupted.
Should I have taken periodic snapshots of the zfs pool it was stored on?
     
mduell
Posting Junkie
Join Date: Oct 2005
Location: Houston, TX
Status: Offline
Reply With Quote
Feb 6, 2012, 08:11 PM
 
Yes, if you would keep snapshots going back far enough to before the sparseimage got corrupted (if it was caused by the application).
     
BLAZE_MkIV
Professional Poster
Join Date: Feb 2000
Location: Nashua NH, USA
Status: Offline
Reply With Quote
Feb 6, 2012, 10:19 PM
 
Then what do you recommend I should do to protect the backup of my backup?
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
Feb 7, 2012, 04:34 AM
 
Originally Posted by CharlesS View Post
Since snapshots aren't coming until the Gold version of ZEVO anyway, why not just wait for that to come out? They have to know that this would be a popular use case, so maybe they'll provide some sort of TM-alike functionality so that you don't have to write that script.
Oh the snapshots are also Gold only? Then I'll definitely wait. I was going to wait anyway because of mirroring support, but this just makes it that much more clear.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
Feb 7, 2012, 04:37 AM
 
Originally Posted by BLAZE_MkIV View Post
Then what do you recommend I should do to protect the backup of my backup?
Since this is the ZFS thread...ZFS mirrors with automatic healing?

What sort of setup do you have today?
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
OreoCookie
Moderator
Join Date: May 2001
Location: Hilbert space
Status: Offline
Reply With Quote
Feb 7, 2012, 06:19 AM
 
Originally Posted by P View Post
Oh the snapshots are also Gold only? Then I'll definitely wait. I was going to wait anyway because of mirroring support, but this just makes it that much more clear.
If you buy the Silver edition now, you can upgrade later for the difference between the two editions. That's why I went for the Silver edition now, but for sure I'll upgrade to Gold at least.
I don't suffer from insanity, I enjoy every minute of it.
     
P
Moderator
Join Date: Apr 2000
Location: Gothenburg, Sweden
Status: Offline
Reply With Quote
Feb 7, 2012, 07:57 AM
 
Originally Posted by mduell View Post
Since it's an SSD I wouldn't be too worried about the impact of write caching off.
That's what I thought at first, but then I remembered all the talk about supercaps and SandForce 1200 vs. 1500 series. There is at least a worry about what happens with a drive when the power goes unexpectedly, so there is some sort of data in DRAM that needs to be written out to flash rapidly in case of power outage.

This makes me wonder how all of this works. The drive has to keep track of a lot more data than just that one block the OS just told it to write. That feature in the end turns into a write cache. So what happens when ZFS tells the SSD to not use it's write cache? Seems to me like we have three options:

1) The drive ignores the command, leaves caches in the default state but telling ZFS that the caches are gone. This means that ZFS doesn't send those regular flush commands. This would actually be the best performance, but the disk reliability goes out the window for the ZIL cache. Since the ZIL cache would get flushed back to the main drive anyway, this is no big deal.

2) ZFS doesn't try to disable the write cache, and sends the flush commands as usual. This means that the HFS+ cache gets flushed all the time, but that doesn't really matter - with the SSD, the bandwidth is there to support something like that.

3) ZFS tries to disable the write cache and performance on the HFS+ partition drops.

3 is really the only problematic setup, so if I can move it to one of the other two manually, that should be enough. Is this setting manually configurable?

Also, for the HDD: The other partition is a bootcamp NTFS partition, ie read only, so ZFS should be able to enable the write cache for that.
The new Mac Pro has up to 30 MB of cache inside the processor itself. That's more than the HD in my first Mac. Somehow I'm still running out of space.
     
BLAZE_MkIV
Professional Poster
Join Date: Feb 2000
Location: Nashua NH, USA
Status: Offline
Reply With Quote
Feb 7, 2012, 11:29 AM
 
Originally Posted by P View Post
Since this is the ZFS thread...ZFS mirrors with automatic healing?

What sort of setup do you have today?
That was mostly a recursion joke.

It's a 4+spare 500gb array in a FreeNAS box. It's a personal machine not work. I've got plenty of open bays but don't really need more storage so I'm waiting for the prices on 2tb enterprise cert drives to drop then I can switch to FreeNAS 8 which at the moment won't import old pools (gui not zfs issue) and move off the aging disks.
     
BLAZE_MkIV
Professional Poster
Join Date: Feb 2000
Location: Nashua NH, USA
Status: Offline
Reply With Quote
Feb 7, 2012, 11:31 AM
 
Originally Posted by P View Post
That's what I thought at first, but then I remembered all the talk about supercaps and SandForce 1200 vs. 1500 series. There is at least a worry about what happens with a drive when the power goes unexpectedly, so there is some sort of data in DRAM that needs to be written out to flash rapidly in case of power outage.
Some of the designs use DRAM as a cache others don't so they don't need the caps.
     
 
Thread Tools
 
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Top
Privacy Policy
All times are GMT -4. The time now is 11:41 PM.
All contents of these forums © 1995-2017 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2017, Jelsoft Enterprises Ltd.,