Re: Problems with BTRFS formatted disk

[Wang Yugui <wangyugui@e16-tech.com>]

Hi,

> On 2022/6/20 09:01, Wang Yugui wrote:
> > Hi,
> >
> >> On 2022/6/20 04:06, David C. Partridge wrote:
> >>> Yes write caching was enabled - I suspect that the way it worked was that on power fail the super-caps held the data until power was restored.
> >>>
> >>> Sadly it wasn't restored for a few weeks by which time the super-caps had lost their charge.
> >>
> >> A little off-topic here, since I'm not familiar with how those hardware
> >> RAID controllers work.
> >>
> >> Yep, those cards should have (super) caps to handle power loss, but for
> >>    SCSI SYNC CACHE commands, they should "the device server ensure that
> >> the specified logical blocks have their most recent data values recorded
> >> in non-volatile cache and/or on the medium."
> >>
> >> Considering in a power loss event, the juice in those caps is definitely
> >> not enough to power those HDDs, it should at least have some
> >> non-volatile cache, like NAND, as backups.
> >>
> >> But from sites I can found, it only states the card has 1024MiB cache
> >> memory.
> >>
> >> Even with caps to keep the memory alive, it's still far from
> >> "non-volatile cache" required by SCSI spec.
> >>
> >> Or is this a common practice in hardware RAID controller world to use
> >> volatile cache and break the SCSI spec requirement?
> >
> >
> > "non-volatile cache"  require a battery(inside or separated) and
> > Backup Flash(inside or separated).
> 
> Then it comes the question, why there is need for battery if our
> non-volatile cache is NAND flash?
> Since NAND doesn't need power to keep its data.
> 
> My guess is, NAND flash is not fast enough so they have battery for RAM
> as the primary cache, and only when power loss happen, then use the
> battery to power the RAM so that we have enough time to dump the content
> of RAM into the backup flash?

Yes.

RAM/memory is fast than NAND. 
and NAND have the limit of total write bytes.