Re: Contention on oraus.msg?

Are you sure it's oraus.msg, not oraus.msb?

I've been playing around tracing an Oracle process on the OS side recently. I noticed that all Oracle processes, background or foreground, open $ORACLE_HOME/rdbms/mesg/oraus.msb (note msb, not msg) with a file descriptor 9 (I only checked 8.1.6 and 8.1.7 for Sparc Solaris 2.6, UFS filesystem). The file is kept open during the lifetime of the Oracle process. Your observation is interesting and I've never seen anyone mention this contention. But posting a little more details should help. What OS or ORA error do you get if any? Can you show us a screen shot with no or very little editing? Note that a file inode is locked when it's being open()ed and read(). I append parentheses even to read because I want to say the inode is locked only when the system call read() (or its variant such as pread, pread64..) is made but released between two system calls. So suppose one process is reading oraus.msb and its trace (truss in Solaris) output shows more than 1 read() line consecutively, in fact between the two read() lines, another process acquires the inode lock and does a read() in the interim. However, since one read of oraus.msb reads 512 bytes, you'll never see 2 consecutive reads of this file because every message in the file is shorter than 512 bytes.

Inode lock during open() can be ignored unless your sessions therefore shadow processes come and go very frequently. Inode lock during read() is inevitable. If this is a serious problem to many people, Oracle has to modify their code to maybe read the message file less often. Knowledge of all this is for informational purpose. There's nothing I can think of you can do to alleviate the problem. Actually, do you really have contention problem on the message file? How many errors do you get and how often? There may be a bug in 8i that almost every SQL command results in reading these words in oraus.msb: "column %s, specified in logmnr.opt file, does not exist in table..."

Yong Huang
yong321_at_yahoo.com

rick_at_oyarsa.com (rlr1031) wrote in message news:<978df3d.0106290244.61c7eaee_at_posting.google.com>...
> Has anyone experienced a problem with contention on the file oraus.msg in a
> high volume, OLTP, OCI-based application? I know the file is used to
> translate messages, but how is it being opened and used by Oracle? And when
> in the processing of an OCI statement is this being hit? On the client? On
> the server? Prior to sending the response?
> Our internal database execution times look good, but we have occasional slow
> downs in the response from the OCI statements.
>
> tia,
> R
> Oracle 8.1.7.x, single Unix platform with clustered disks (sorry cannot be
> more specific at this time)
Received on Fri Jun 29 2001 - 12:36:09 CDT