|
> The biggest gotcha I can see is that you're trying to replicate data
> in a 'bottom up' fashion using this whereas RAID solutions expect to
> work 'top down'.
>
> How do you get the RAID system to recognise changes in the netwrok
> block device, other than by forcing the local device to resilver?
Well, the way I was intending to use this was with a "main" server
being read/write, and the slaves being read only. Thus, this should never
become a consideration -- the slave servers will never change, except on
command from the master.
> AFAICT, that has to be a very manual process. Wouldn't it be just as
> easy to push-mirror off the master server in usersapce using something
> like rsync?
Well, if you let RAID handle it automatically, it's, well... automated.
You don't have to make anyone do anything -- it just happens. I
personally don't have a problem with using rsync/rdist/etc, but any time
you can automate anything (safely), it's a good thing.
Thanks,
Kyle Sparger
On 14 Apr 2000, Stephen Zander wrote:
> >>>>> "Kyle" == Kyle Sparger <ksparger@xxxxxxxxxxxxxxxxxxxx> writes:
> Kyle> Once you can export a block device over the network, you can
> Kyle> include that device in a software RAID-1 array, which will
> Kyle> handle all the replication for you.
>
> The biggest gotcha I can see is that you're trying to replicate data
> in a 'bottom up' fashion using this whereas RAID solutions expect to
> work 'top down'.
>
> How do you get the RAID system to recognise changes in the netwrok
> block device, other than by forcing the local device to resilver?
> AFAICT, that has to be a very manual process. Wouldn't it be just as
> easy to push-mirror off the master server in usersapce using something
> like rsync?
>
> --
> Stephen
>
> "If I claimed I was emporer just cause some moistened bint lobbed a
> scimitar at me they'd put me away"
>
|
