"truss" on Solaris and AIX, "tusc" on HP-UX, and
"strace" on Linux...
----- Original Message -----
Sent: Tuesday, April 30, 2002 4:10
PM
Subject: Re: FILE IDENTIFY wait
event
Depending on the OS, one could maybe trace the call made by
Oracle's OSD-layer to the OS (I can't remember the name of the trace facility
in Solaris, but it's also sort of available now in AIX version 5 and there are
similar freeware things available for HP/UX and other OS'es).
Tim
Gorman wrote:
If a wait-event (i.e. FILE IDENTIFY) was
posted, then Oracle was undoubtedly making a system call such as "open()" or
"fstat()" or something. If those system calls "hang" when the device
file exists but the underlying device is offline, then there is nothing
Oracle can do about it...
It could be platform-specific behavior;
it might "hang" on one platform and fail on another...
-----
Original Message -----
Sent:
Tuesday, April 30, 2002 11:48 AM
Subject:
RE: FILE IDENTIFY wait event
The volume was offline. Once it was back online,
Oracle resumed the process of creating a tablespace.
Which brings up another question...why did Oracle
wait on this condition instead of terminating with an error? If
a file-system directory/drive cannot be found, the attempt to create
a tablespace will fail. Is there something different in using a raw
device?
Dan
Make sure that the volume group underlying
the logical volume is ONLINE?
To test, I like to do something quick like
"dd if=<raw-device> of=/dev/null count=100" to test the
readability of the raw device. If this fails, you can go to the
SysAdmin with something that is pure UNIX...
-----
Original Message -----
Sent:
Tuesday, April 30, 2002 10:41 AM
Subject:
FILE IDENTIFY wait event
Oracle8.1.7 on Solaris8
I am
trying to create a tablespace on a raw device and the SQL*Plus session
that performs the tablespace creation is waiting on a 'file identify'.
A quick search on metalink and other docs did not indicate the root
cause of the wait event.
I have
used the raw device previously and the create tablespace uses the
'REUSE' option.
Okay
gurus...any clues as to exactly what conditions cause the
event?
Dan
Fink
|