[DRBD-user] DRBD Question

paddy at panici.net paddy at panici.net
Wed Jan 10 12:15:36 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 Tue, Jan 09, 2007 at 04:15:00PM -0500, Brad Barnett wrote:
>
> If you are sure one machine would be able to take both loads, your
> argument is not logical.  It does not matter, economically, if you use one
> machine and have another acting as purely a drbd slave.  You have not
> "saved" anything.  You still have two machines.  

Not all failures are crises, some are graceful.

an active/active setup, provisioned such that a single node can take
the full load in the case of one failure, can still (in principle)
be a gain from an active/passive configuration if it simply
delivers *better* performance (for whatever value of "better
performance" you prefer).

Whether that will apply in practice will also depend on the problem
domain: some problems lend themselves more easily to paralellisation,
others do not.

In any case, the active/active configuration will generally be more
complex, and the gain probably slight. So it is certainly good
advice to be wary of confusing worry about redundant capacity
going to waste with the possibility of optimising the use of
redundant capacity (at the likely cost of additional complexity).

then there is the n+1 instead of 1+1 redundancy kind of thing ...

Regards,
Paddy



More information about the drbd-user mailing list