On 14 November, 2012 - Michel Jansens sent me these 1,0K bytes:

> Hi,
>
> I've ordered a new server with:
> - 4x600GB Toshiba 10K SAS2 Disks
> - 2x100GB OCZ DENEVA 2R SYNC eMLC SATA (no expander so I hope no SAS/ 
> SATA problems). Specs: 
> http://www.oczenterprise.com/ssd-products/deneva-2-r-sata-6g-2.5-emlc.html
>
> I want to use the 2 OCZ SSDs as mirrored intent log devices, but as the 
> intent log needs quite a small amount of the disks (10GB?), I was  
> wondering if I can use the rest of the disks as L2ARC?
>
> I have a few questions about this:
>
> -Is 10GB enough for a log device?

Our log device for the department file server, for roughly 100
workstations etc, seems to hover about 2MB used. It's only sync writes
that goes here, and it's emptied at the next transaction.

So check how much sync writes you have per flush (normally 5 seconds
nowadays, used to be 30 I think?). If you are pushing more than 2GB of
sync operations per second, then I think you should get something
beefier ;)

> -Can I partition the disks (10GB + 90 GB) and use the unused (90GB)  
> space as L2ARC?
> -If I use the rest of the disks as L2ARC, do I have to mirror the L2ARC 
> or can I just add 2 partitions (eg: 2 x 90GB = 180GB)
> -If I used non mirrored L2ARC, Could something go wrong if one L2ARC  
> device failed (pool unavailable,lock in the kernel, panic,...)?

It's checksummed and verified, shouldn't be a problem even if it fails
(could be a problem if it's half-failing and just being slow, if so -
get rid of it).

>
>
> --
> Michel Jansens
>
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss@opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss


/Tomas
-- 
Tomas Forsman, st...@acc.umu.se, http://www.acc.umu.se/~stric/
|- Student at Computing Science, University of Umeå
`- Sysadmin at {cs,acc}.umu.se
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Reply via email to