performance of mirror and re-sync

[Klorese, Roger B.A.]

For a mirrored device, we need to guarantee that as soon as we make it
available to you, the plexes are identical.  (If we were not to sync,
you could run a program which opens the raw volume and reads a block,
and run it twice and get different data -- a real non-no.)  The
underlying mechanism can't really tell if you're adding a mirror to a
junk volume you're about to initialize both plexes of (your case), or if
you're adding a mirror to an encapsulated volume with good data.

The new Java GUI will, either in the first release or soon after, allow
you to say that the mirror has uninitialized data in it, and will skip
the mirror-syncing step.

>-----Original Message-----
>From:	dck@mink.mt.att.com [SMTP:dck@mink.mt.att.com]
>Sent:	Thursday, March 12, 1998 11:08 AM
>To:	ssa-managers@Eng.Auburn.EDU
>Subject:	performance of mirror and re-sync
>
>Hi, you'll,
>
>I noticed that while setting up mirrors using a 2nd SSA to mirror the
>corresponding disk on the 1st SSA (RAID-1, not stripping or 5), that adding
>the mirror takes like 5-10 mins/disk even if the filesystme on it is
>empty.  Why is that?  There is no data to sync up.
>
>David C. Kwong
>dck@mink.att.com