Journaling UFS with gjournal.

[Oliver Fromme]

Eric Anderson <anderson at centtech.com> wrote:
 > Oliver Fromme wrote:
 > > [...]
 > > However, if the size of the journals provider is twice the
 > > maximum raw transfer rate of the disk multiplied by the

I should have been more specific:  "of the disk containing
the _journal_ provider", of course.

 > > switch interval time (as you recommended), it should be
 > > impossible to get a journal overrun.
 > 
 > What if the journal had to write out little bits across the entire 
 > journaled area, so the throughput was greatly reduced, while the 
 > journaled area had massive heavy random reads keeping the disk heads 
 > very busy?

In that case there is even less danger of an overrun,
because the many seek operations of the disk will reduce
the throughput, so it will be far from reaching the end
of the journal space before then end of the journal switch
interval.

Pawel, please correct me if I'm wrong.

The maximum raw transfer rate of the journal provider is a
hard limit which cannot be exceeded.  If you configure the
size of the journal provider to be twice that maximum rate
multiplied by the switch interval, there is simply no way
to hit the end of the journal space before the end of the
switch interval.

The worst case would happen when there have been no write
operations for a while, so the journal is empty, and the
inactive journal does not need to be read (i.e. the disk
is idle).  Now a huge amount of write operations start,
causing to fill the active journal sequentially with the
maximum possible throughput (almost no seek operations of
the disk heads).  The recommendation for the size of the
journal provider should account for that worst case.

Best regards
   Oliver

-- 
Oliver Fromme,  secnetix GmbH & Co. KG, Marktplatz 29, 85567 Grafing
Dienstleistungen mit Schwerpunkt FreeBSD: http://www.secnetix.de/bsd
Any opinions expressed in this message may be personal to the author
and may not necessarily reflect the opinions of secnetix in any way.

"One of the main causes of the fall of the Roman Empire was that,
lacking zero, they had no way to indicate successful termination
of their C programs."
        -- Robert Firth