[DRBD-user] DRBD and write cache

Luciano Miguel Ferreira Rocha strange at nsk.no-ip.org
Wed Mar 14 11:53:41 CET 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Wed, Mar 14, 2007 at 07:28:22AM +0200, Siim Vahtre wrote:
> Thank you for your answer!
> 
> >if your hardware pretends to have done something which it actually did 
> >not do, there is no drbd magic to cover for you.
> 
> I am completely aware of that and I was not hoping for any "magic". I just 
> thought that maybe thanks do ordered writes and some consistency 
> flags/counters,

But your controller doesn't guarantee that, as it can reorder internally
the requests (hey, if it doesn't obey flush requests, why would it obey
the order of the requests when finally flushing?).

Note that your controller is broken as far as any transactional system is
concerned. Postresql, Oracle, etc. *will* believe it when it says it
committed some data to persistent storage, so if your controller or
power decides to die then, you've just lost a bunch of transactions your
system believed commited. In short, there's no ACID.

Some controllers that provide write-cache also support an internal
battery that provides power long enough to flush the data. Check your
manuals or vendor.

-- 
lfr
0/0
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20070314/6249ef0f/attachment.pgp>


More information about the drbd-user mailing list