Resending because original got too long with all the quoted replies. On Mon, May 16, 2011 at 09:51, Charles Schultz <sacrophyte@xxxxxxxxx> wrote: > Yes, it does. Touching the alert.log has no impact on whether or not the > database can write to it (in our case). In a theoretical situation, it > depends on what exactly is open and when it was opened. :) > > Some new developements in my specific case. > I did a truss -aefo on sqlplus, executed dbms_system.ksdwrt (I used both 3 > and 2 for kicks), and the output shows one and only one line pertaining to > the alert.log: > access("./alert_TEMDEV.log", F_OK) = 0 > > No open, no write, no read, no nothing else. And why ./ ? I have only two > files with the same name on my entire disk, and neither of them were updated > today, nor have the output from the ksdwrt command. It would seem as if my > database has some memory of alert_TEMDEV.log in its brain somewhere, but it > cannot recall exactly where it is, nor what it is supposed to do with it. As > I have mentioned previously, I have scoured all the file descriptors in > /proc/*/fd and have not found any files that should have the ksdwrt output > other than log.xml. Which tells me that no process is actually even > attempting to write to alert_TEMDEV.log. Which is quite odd. > > > On Mon, May 16, 2011 at 09:36, Mark W. Farnham <mwf@xxxxxxxx> wrote: > >> Doesn’t an rm’d file become persistent but become invisible by name to >> file system services (and available only to the processes that already have >> it open) until all those processes end or close it? >> >> >> >> What happens if you touch alert.log in the location where alert.log is >> supposed to reside? >> >> >> >> When you bounced the database, did you exit the session where you did the >> shutdown, or was that shutdown and then startup from within the same >> session? >> >> >> >> alert.log and xml.log are hammering the same place as per Oracle current >> default “best practice” to make it easy to find all logs by smashing the >> same place with all of them, right? So that pretty much exonerates it being >> a device or permissions problem. >> >> >> >> Regards, >> >> >> >> mwf >> > -- Charles Schultz