iSCSI + VM + static Block size?

[BERKUT]

Hello.
First, sorry for my not good english or noob question :)

At this moment we have this configuration.
6 x 2 TB hdd in mirror ~ 6 TB free space - 50% for iSCSI = 3 TB free space for VHD.
Lost about 75% free space...

So, if we disable "Dynamic Record Size" and set static (for example 8K block size). We can use without problem iSCSI over 50% for Virtual Machine hard drivers?
Thank you.

 

[jgreco]

Hello.
First, sorry for my not good english or noob question :)

At this moment we have this configuration.
6 x 2 TB hdd in mirror ~ 6 TB free space - 50% for iSCSI = 3 TB free space for VHD.

Ok.

Lost about 75% free space...

No, lost about 75% RAW space. Only lost 50% free space.

So, if we disable "Dynamic Record Size" and set static (for example 8K block size). We can use without problem iSCSI over 50% for Virtual Machine hard drivers?

You can use as much as you want, but the normal experience is that the fuller a pool gets, and the more writes a pool gets, the harder fragmentation hits you. We typically find that people filling a pool more than 50% eventually run into pain if they're doing normal amounts of writes to their VM's. However, there's nothing magic about "50%". You can get ZFS pretty screwed up with fragmentation on a system that's only 20% capacity. This is kind of independent of any choices you make about block size.

The usual fix is to be aware that, from time to time, you might need to shut down some of your heavy writing VM's and "move" them off the pool and then back onto the pool.

 

[BERKUT]

Hm, ZIL (SSD of course) can be good choice if pool go >50%?
If I understand correctly, VM write - > ZIL - > zpool.

 

[jgreco]

You misunderstand.

https://forums.freenas.org/index.php?threads/some-insights-into-slog-zil-with-zfs-on-freenas.13633/

Code:

VM write -----------> ZIL
        \----> txg -> pool