Last weekend I performed a Finder copy of pool filesystems from one pool to another. Both pools were connected to the same host via SAS and SATA.
10.10.2
ZFS 1.3.1 RC5
GB X58 UD7
Xeon QC @ 3.2Ghz
24GB DDR1600MHz
RocketRaid 2744 PCIe 16X Lane HBA (set to JBOD), (6 x 4TB disks)
Addonics 6G 10-port SATA PCIe 2X Lane HBA (non RAID), (8 x 2TB disks)
I've since disconnected the 2nd pool and connected it up to an xServe, so the hostnames differ accordingly in the pasted output below.
setup:
- Code: Select all
HacProServer:~ madmin$ zpool status
pool: ztank
state: ONLINE
scrub: scrub completed after 5h14m with 0 errors on Sun Mar 15 14:40:38 2015
config:
NAME STATE READ WRITE CKSUM
ztank ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
disk6 ONLINE 0 0 0
disk5 ONLINE 0 0 0
mirror-1 ONLINE 0 0 0
disk4 ONLINE 0 0 0
disk3 ONLINE 0 0 0
mirror-2 ONLINE 0 0 0
disk2 ONLINE 0 0 0
disk1 ONLINE 0 0 0
errors: No known data errors
HacProServer:~ madmin$
The above is using 4TB drives and below is using 2TB disks
- Code: Select all
Xserve:~ badmin$ zpool status
pool: btank
state: ONLINE
scrub: scrub completed after 9h16m with 0 errors on Sun Feb 22 23:06:44 2015
config:
NAME STATE READ WRITE CKSUM
ztank ONLINE 0 0 0
raidz-0 ONLINE 0 0 0
disk10 ONLINE 0 0 0
disk9 ONLINE 0 0 0
disk8 ONLINE 0 0 0
disk7 ONLINE 0 0 0
raidz-1 ONLINE 0 0 0
disk14 ONLINE 0 0 0
disk13 ONLINE 0 0 0
disk12 ONLINE 0 0 0
disk11 ONLINE 0 0 0
errors: No known data errors
Xserve:~ badmin$
filesystems on ztank are:
- Code: Select all
madmin$ zfs list
NAME USED AVAIL REFER MOUNTPOINT
ztank 7.63T 3.08T 460K /Volumes/ztank
ztank/Comedy 201G 3.08T 201G /Volumes/ztank/Comedy
ztank/Docos 749G 3.08T 749G /Volumes/ztank/Docos
ztank/Files 74.6G 3.08T 74.6G /Volumes/ztank/Files
ztank/Movies 2.53T 3.08T 2.53T /Volumes/ztank/Movies
ztank/Music 155G 3.08T 155G /Volumes/ztank/Music
ztank/Pictures 154G 3.08T 154G /Volumes/ztank/Pictures
ztank/Sport 119G 3.08T 119G /Volumes/ztank/Sport
ztank/TVShows 2.65T 3.08T 2.65T /Volumes/ztank/TVShows
ztank/Torrents 933G 3.08T 933G /Volumes/ztank/Torrents
ztank/Video 120G 3.08T 120G /Volumes/ztank/Video
HacProServer:~ madmin$
the btank pool has exactly the same filesystems, just an underscore at the start of each of the names to get around the Finder bug of two filesystems of same name but on different pools.
therefore
- Code: Select all
NAME
btank
btank/_Comedy
btank/_Docos
btank/_Files
btank/_Movies
btank/_Music
btank/_Pictures
btank/_Sport
btank/_TVShows
btank/_Torrents
btank/_Video
I started with the Comedy filesystem and it transferred over in about 20mins, and so I started with the next file system being Docos.
As it was late in the night I let it run during the night and intended to inspect it after some shut-eye.
In the morning, I switched the monitor back on and the system was totally locked up and frozen screen. The time in the menubar indicated that it happened about 30mins after I kicked off the transfer.
After rebooting, destroying the _Docos filesystem and recreating it and kicking off the transfer, sure enough it locked up again.
I tried another filesystem being Movies. it locked up too. I tried a smaller filesystem such as Video, it copied over fine, as did Sport, Pictures and Music.
So it seemed that small filesystems were transferring okay.
What I did notice is that MenuMeters in the menu bar was at times showing that the system only had 54MB out of 24GB free! This scared me a bit but I figured ZFS should release memory appropriately when he OS wants it back.
Not being 100% confident that ZFS was releasing memory fast enough, before kicking off another Finder copy of the larger filesystems, I did some homework and found the kstat.zfs.darwin.tunable.zfs_arc_max command for RC5 so I set it to use 16GB using:
- Code: Select all
sudo sysctl -w kstat.zfs.darwin.tunable.zfs_arc_max=17179869184
and
- Code: Select all
$ sudo sysctl -w kstat.spl.misc.spl_misc.simulate_pressure=<amount of ram to release in bytes>
to release the memory immediately.
I kicked off the Finder copy of ztank/Docos to btank/_Docos and it successfully transferred, as did the remaining larger filesystems of TVShows, Movies and Torrents.
I was watching arcstat.pl during this and everything seemed to behave itself.
So, I'm curious if anyone else has encountered this when using the Finder to copy over large filesystems?
In case you're wondering, this backup is my 2nd backup which is why I don't use zfs send/recv (plus I was lazy) and only do this backup about every 90 days.
Regards,