On Tue, 26 Jun 2001, Jesse, Rich wrote:
> First, I believe it's a misconception that on a Unix system there can be no
> data lost in an Oracle DB from a system crash. This HAS to be a function of
> "syncer", doesn't it? And, therefore, until syncer decides any buffer
> writes actually go to disk, transactions can be toast. Granted, this is a
> very short time, but the possibility would still exist for a standalone
> Oracle DB, especially for one with a high transaction count. But I haven't
> seen any "official" info, whether true or false, from Oracle about this.
The question of whether a write results in data being written directly to
disk or only to the UBC depends on the "oflag" that is used when a process
open a file with the open() call.
Oracle's log writer (I believe) opens the online redologs with the O_DSYNC
flag. That means that a given write() call will not return successfully
(and thus a commit will not return as complete) until that data has been
written down to disk. Syncer is only responsible for the data's write out
to disk if a flag like O_RDWR was used to open the file.
Your concern may apply to a disk subsystem and controller that have their
own cache. In the case of these systems, they return success to the
operating system as soon as a write has been successfully added to the
controller's cache. Such controllers have battery backups, so that in case
power is lost to the unit, pending writes in the cache can complete before
the unit powers off. That battery backup on caching disk controllers is
pretty key to running Oracle successfully.
This is just the kind of caveat that I think would be good to include in a
presentation on misconceptions.
> Second, I hope you're going to have explanations and/or qualifications (even
> brief ones!) about the misconceptions somewhere on your website? There's a
> few in your list that have me intrigued!
Of course I plan to explain them, but tell me what intrigues you. I'd love
to hash the issues out here before I make a fool of myself in SF in
December.
--
Jeremiah Wilton
http://www.speakeasy.net/~jwilton
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Jeremiah Wilton
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
