On 2015-12-02 18:10, Karel Gardas wrote:
On Wed, Dec 2, 2015 at 10:39 AM, Tinker <ti...@openmailbox.org> wrote:
On 2015-12-02 17:31, Karel Gardas wrote:
I think RAID1C is capable of detecting mis-directed write. I wrote
about it in some of my previous email.
Hi Karel,
I'll follow up on the other things in a separate email later, but, you
clarify that you think RAID1C has protection against misdirected
writes
already - I don't understand how it works, can you please explain to
me
*exactly* how it can be said to be solid against misdirected writes?:
Let's assume the following nightmare scenario:
* You write data to sector X. All the physical writes for that on all
the
underlying disks are mis-directed (about 4 writes).
I've been talking about mis-directed write (singular!) while you
suddenly switch to talk about writes (plural!). So what you describe
here is basically you get *all* writes to *all* drives mis-directed?
Oh, what's the probability of this? Well, if your read on the drives
is not mis-directed in the same way (why would we then be talking
about mis-directed write right?), then you read old data from the
sector silently. BOOOM! :-)
But allow me to counter with the question: what will your scheme do if
*all* writes to *all* your drives are mis-directed? Wouldn't it also
return old data on read from X? :-)
Karel,
To answer your question: In that case, as soon as that invalid data
would actually be read from disk, it would be caught by the checksums
that are guaranteed to be kept in RAM, so that is, the first-level
checkums (or the über-checksum) match would fail.
So that's what's nice about having all the disk checksummed, that data
security works even if all writes fail. That moves the risk surface
altogether from the whole time period when the disk is in use, to only
the time between being taken out of use and being taken in use again.
So that's the difference here.
Thoughts?
Btw, any checksum algorithm would work for implementing a tree like this
by the way, even CRC64 I guess. So Fletcher as such is out the window. I
intend to followup on your other emails in some hours.
Tinker
