Page 1 of 2

If your hackintosh is configured properly..

PostPosted: Sat Mar 15, 2014 6:13 am
by dmilith
If your hackintosh is configured properly - ZFS will work properly.
If you want to know my Hackintosh spec, it's mentioned here: http://www.tonymacx86.com/members/dmilith/

What just works:
* zpool v28++ 5000, zfs file system version 5
* deduplication
* compression (I use lz4)
* sleep
* auto loading kexts with system

What doesn't:
* auto import for ZFS pools. It must be done manually after each reboot (for example using Lingon and launchdscript)
* it seems that kernel panic dumps aren't created on Hackintoshes like on native Macs. Adding "-v" to kernel_flags might help.
* TRIM support (not yet implemented like many other features)
* reloading zfs/spl kexts with mounted ZFS pools (hello captain Obvious!)
* mounting ZFS pools from DiskUtility and family.


Hacks and tricks:
* NVRAM feature works a bit differently than on native Macs: For example to enable kernel debugger in case of panic do: sudo nvram boot-args="debug=0x144"


TADA!
;]


FreeNode: #openzfs-osx

Re: If your hackintosh is configured properly..

PostPosted: Fri Apr 04, 2014 6:58 am
by modahamburger
Hi,

also running OpenZFS on a hackint0sh. Works like a charm, also auto mounting at boot.

Have a look here: https://openzfsonosx.org/wiki/Autoimport

Re: If your hackintosh is configured properly..

PostPosted: Sun Apr 13, 2014 6:38 am
by TomSoutherland
Yep... works like a champ. I've written a short document covering my conversion of /Users to zfs. Check it out at http://webcave.us/doku.php/prophead/mac_tips/zfs if interested

Re: If your hackintosh is configured properly..

PostPosted: Wed Apr 16, 2014 6:36 pm
by AlexWasserman
Tom,

Interesting write-up. I do much the same - couple of things I've done that can help.

1. Create a spare admin user with a non-ZFS home directory. If you right-click on a user in the Users Sys Prefs pane you can access the advanced options and put their home directory somewhere other than /Users. This is handy for those times when ZFS hasn't, or won't mount, as it means you don't have to go through the new-user dialogs, and the system won't put new users directories into /Users on your boot drive.

2. Not sure why you created different filesystems for each user. I created a single filesystem for /Users, and have other filesystems for various different things. Shared media (movies, etc) go under /Media, without compression as it doesn't do much on already compressed files, for example. You can then set your root ZFS filesystem not to mount, and just mount your sub-filesystems wherever you want.

Re: If your hackintosh is configured properly..

PostPosted: Fri Apr 18, 2014 7:05 am
by TomSoutherland
Good call on the spare admin account. That's a must come upgrade time.

As for the seperate filesystems... I want the flexibility to choose what filesystem to replicate (send/receive) and where they are replicated. It does cause a bit of clutter in the finder but I can live with that. Thanks for the feedback and hints.

Re: If your hackintosh is configured properly..

PostPosted: Sat Apr 19, 2014 5:51 pm
by emory
AlexWasserman wrote:As for the seperate filesystems... I want the flexibility to choose what filesystem to replicate (send/receive) and where they are replicated. It does cause a bit of clutter in the finder but I can live with that. Thanks for the feedback and hints.


You can always drag the extraneous filesystems out of 'Devices' in the Finder sidebar. (That's what I do. I haven't investigated setting that as a default preference for other users but my Hac is used by me 99% of the time anyway.)

Re: If your hackintosh is configured properly..

PostPosted: Sat Apr 19, 2014 10:43 pm
by ilovezfs
emory wrote:You can always drag the extraneous filesystems out of 'Devices' in the Finder sidebar. (That's what I do. I haven't investigated setting that as a default preference for other users but my Hac is used by me 99% of the time anyway.)

You can use the com.apple.browse property for this.

Code: Select all
zfs set com.apple.browse=off foo/bar

Re: If your hackintosh is configured properly..

PostPosted: Mon Apr 21, 2014 5:45 am
by TomSoutherland
Excellent tip... Thanks!

Re: If your hackintosh is configured properly..

PostPosted: Sat Dec 06, 2014 5:37 pm
by lundman
The ZFS code in 10a286 is what MacZFS was based on, and the glue magic it contains are in O3X SPL layer. But the ZFS code is a different beast since pool v8. But we are getting much better. Might not have hundreds of days of uptime yet (we will, I love uptime) but so many more features. We can now use more than 4GB of ARC, lz4 compression and all those goodies.

Re: If your hackintosh is configured properly..

PostPosted: Sun Dec 07, 2014 3:51 pm
by lundman
dragonmel wrote:
2> Spotlight reindexed I think, even though they should have had good spotlight data files at the root of each volume in the pool already?. When booting back into 10.6.7 it reindexed again... is that spotlight being different between 10.6 and 10.10 or is it openzfs doing that?


It will do that. Spotlight stores the last-unmount-time so it knows if it needs to rescan. MacZFS stores this in a root xattr. We store it in a dataset value. So going between the versions will cause a rescan.


4> when I booted back into 10.6.7 it auto mounted my pool but it lost its zpool history? completely reset? I did not upgrade the pool or anything. Ideally I would have preferred to import it the first time as read only but it didnt give me the option... my recommendation is if a pool is exported it should not be auto imported the next time. it should only be auto imported if it was previously imported and mounted from the previous boot. ... but only if that is standard ZFS behavior since we want to stay compatible at the ZFS level.

I don't know about history, in theory neither version should remove history, but they did re-do how history works somewhere between pool v8 and v28.
I don't know how autoimport works in either versions, someone else will have to comment :)


5> I have 8gb of ram and no dedicated ssds since 10a286 didnt support it. Obviously I am not going to be able to run de-dupe.. should I make any other changes to settings to take advantage of features or just use ashift=12 since I am running 2TB drives and let ZFS set arc etc. When I do decided to switch over.. I will need to destroy my pool and rebuild it using ashift since I didn't before and then re-import from my backup. I have been using copy software to do my backups because it was keeping permissions and custom folder icons etc... would moving a snapshot be faster to copy my entire pool to the new set?


No dedup, ashift yes, and I would use compression=lz4. Snapshot with zfs send would be the best way to move the data. But rsync will suffice if you get the options right.


I have been running 10a286 on that machine since 10.6.. a long time... no data loss and TBs worth of data and media stored without issue.... is 1.3.1-rc2 stable enough to run on or are there crashes and data losses being reported?


No data loss, but its possible you will have deadlock or panic until they are remove. There is a plan for rc3 to come, but nothing too major in rc2.

if you can get in hooked back into the operating system and core storage a little tighter like it used to be.. I would be thrilled!!


The now famous issue116. It is coming.