Is this correctly configured?

[jschmok]

Hi All, I'm a FreeNAS newb, but have spent many hours poking around on it.... its starting to play nice. Anyways, I just checked my Volume Status and wanted to get someone with more experience to confirm if this is indeed a RAID10 or Striped Mirrors in ZFS Terminology.

I have an SSD drive as L2ARC as well

This is running version 11.1-U6

Thanks!

 

[gpsguy]

Yes, your pool consists of striped mirrors.

Striped Mirrors in ZFS Terminology

 

[jschmok]

Thanks! It's just confusing because it says 'stripe' at the bottom for the cache drive... Not above the actual drives....

 

[Arwen]

L2ARC drives, (aka Cache), are always striped, as mirroring them is silly. (The original data is on the pool disks, so any disk loss is not fatal.) Since you have only 1, the wording can look odd unless you have it explained.

Now, SLOG, (aka Log, or separate ZIL), can be mirrored. And should be under some circumstances.

 

[jschmok]

What exactly is SLOG ? Is it simply a log of transactions? What effect does it have on performance?

 

[Arwen]

For datasets or Zvols that use "sync=always", writes are written to the Separate ZFS Intent Log first, then the write is acknowleged to the writer task. In time, the memory copy of the data is pushed to the data disks in the pool. Thus a SLOG is write mostly, and generally never read. (So a high endurance, with high speed writes AND power loss protection SSD is recommended for SLOGs.)

If a crash occurs after the write to the SLOG, but before the data is pushed to the data disks in the pool, the SLOG has a copy of the data. So on power up, ZFS will flush the data in the SLOG to the data disks.

Basically a SLOG is the opposite of a read mostly Cache device. But, a SLOG is really only needed when using a dataset or Zvol for VM storage.

 

[pro lamer]

For datasets or Zvols that use "sync=always", writes are written to the Separate ZFS Intent Log first

@jschmok
IIRC if a dataset doesn't have a SLOG but uses "sync=always" the ZIL (ZFS Intent Log) is located on spinning rust... (just a detail chime in) Thus it is then not named SLOG because it is not separate. Nice reading here (Slideshow explaining VDev, zpool, ZIL and L2ARC).

 

[Arwen]

@pro lamer, you are correct.

 

[RickH]

Thanks! It's just confusing because it says 'stripe' at the bottom for the cache drive... Not above the actual drives....

A zfs pool will always 'stripe' the data across the individual vdevs (this is why there's no recovery for pools that have completely lost a vdev) - the drive redundancy is handled at the vdev layer. The vdevs themselves can be any mix of mirrors, raidZ1, Z2, or Z3 and can be comprised of different sized disks (there are definitely some 'best-practices' that come into play here about keeping things more uniform, but mixing all these different types will work)

In your screenshot 'mirror-0' and 'mirror-1' are your 2 vdevs that are each comprised of 2 disks in a mirrored config. Your pool 'jedi_archives' is striping the data across both available vdevs. In the future it would be possible to add 2 additional disks to a 3rd mirrored vdev and expand your pool size (assuming you have the room in your case, ports, etc...)