ZFS on Linux 0.8.0 released!

[BLinux]

Looks like a lot of features and enhancements in this one:

zfsonlinux/zfs

direct I/O no longer gives an error:

$ dd status=progress if=/dev/zero of=speedtest.out bs=1M count=10240 oflag=direct
8667529216 bytes (8.7 GB, 8.1 GiB) copied, 5 s, 1.7 GB/s
10240+0 records in
10240+0 records out
10737418240 bytes (11 GB, 10 GiB) copied, 5.67395 s, 1.9 GB/s

in addition:

+ native encryption (AES-256-CCM)
+ device removal
+ pool checkpoints
+ TRIM for SSDs
+ pyzfs for Python interface
+ allocation classes
+ QAT acceleration
... and more...

trying it out on my workstation first before I even try it on my servers. must upgrade existing pools after a reboot to enable the new features.

 

[Brian Puccio]

Scrubbing and resilvering is supposedly significantly faster now. Trying to dig up some numbers on the performance boost.

 

[optimans]

Just be weary that the latest kernel updates 4.14.x, 4.19.x, 5.x no longer have SIMD acceleration.

NixOS has a patch for it; wonder if any other distro's will do the same.

 

[Rand__]

What about sync write speeds? Thats *the* one thing keeping me from looking into ZoL at this time...

 

[elag]

I now have a 6disk RAIDZ2 on Freebsd 12. Is this the right moment to move to ZOL on Centos 7? Will I be able to import the existig pool? If so I was thinking of giving it a try. The reason for the move would be: on-board support for freeipa integration for NFSv4 as well as standardization on a single platform. Will this work?

 

[optimans]

There is a benchmark report on the new FreeBSD with iX ZoL (12.0-STABLE) that shows higher IOPS than normal FreeBSD with Illumos ZFS.

@elag I would test the iX version and see how that goes? Make sure to backup everything first just in case

FreeBSD 12-Stable
Index of /62993/trueos/iso/freebsd12-zol/

FreeBSD HEAD
Index of /62993/trueos/iso/freebsd13-zol/

 

[TeeJayHoward]

+ device removal

I was confused on this point (Can't we already remove/replace faulty drives?), so I clicked the link and read up.

• Device removal #6900 - This feature allows single and mirrored top-level devices to be removed from the storage pool with zpool remove. All data is copied in the background to the remaining top-level devices and the pool capacity is reduced accordingly.

If I'm understanding it correctly, we can now non-destructively shrink pools. If so, that's incredible!

Playing with it now on my backup server. It looks like it's still in testing for RHEL/CentOS:

yum --enablerepo=zfs-testing install kernel-devel zfs

edit: Looks like you'll need to reboot after the install in order to upgrade the pools.

[root@backup ~]# zpool upgrade
This system supports ZFS pool feature flags.

All pools are formatted using feature flags.


Some supported features are not enabled on the following pools. Once a
feature is enabled the pool may become incompatible with software
that does not support the feature. See zpool-features(5) for details.

POOL  FEATURE
---------------
backup
      encryption
      project_quota
      device_removal
      obsolete_counts
      zpool_checkpoint
      spacemap_v2
      allocation_classes
      resilver_defer

[root@backup ~]# zpool upgrade backup
This system supports ZFS pool feature flags.

cannot set property for 'backup': invalid feature 'encryption'

edit2: Oh, boo. "Top level devices" means vdevs. I can't remove a single disk from a pair of raidz3s.

 

[kroy]

I was confused on this point (Can't we already remove/replace faulty drives?), so I clicked the link and read up.

• Device removal #6900 - This feature allows single and mirrored top-level devices to be removed from the storage pool with zpool remove. All data is copied in the background to the remaining top-level devices and the pool capacity is reduced accordingly.

...

edit2: Oh, boo. "Top level devices" means vdevs. I can't remove a single disk from a pair of raidz3s.

I've actually been using this for a while (moved a very large pool from OmniOS CE when it got this ability last year).

I laid out a bunch of tests that should make it clear here