Anyone whose ever had to rebuild a server just fromtapes might agree

John Macdonald jmm-TU2q2He6PgRlD5gtYiU6kEEOCMrvLtNR at public.gmane.org
Tue Jan 6 07:38:00 UTC 2004


On Mon, Jan 05, 2004 at 10:52:04PM -0500, cbbrowne-HInyCGIudOg at public.gmane.org wrote:
> The only relatively "safe" way to get a quasi-atomic backup is to use
> LVM, so that you can atomically "split off" a temporary copy of the
> filessystem.  That makes the backup "atomic," which can allow satisfying
> databases' needs for consistency.  You may need to recover the database,
> but that usually fits into its design.

With IBM's DB2 UDB database, there is an improvement
to be made to this process.  It has a command to
suspend write IO (which also updates the log before
stopping writes).  Then you can split off a copy,
and resume writes.  The copy can then be restarted (no
recovery required).  Since it is only writes that are
suspended, read transactions are not blocked at all,
and write transactions can be carried out up to the
point of committing (which then has to wait for the
write resume).  Depending upon the mechanism used to
split off the copy, this can be a very fast operation
- using EMC Symmetrix storage array, I split off a
copy of a 64 MB database with writes suspended for
only a few seconds.
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml





More information about the Legacy mailing list