You may not have exited from the session that was used to shutdown one of those databases. If so, the corresponding IPC resources would remain marked for deletion, but not yet free for reuse.

Another possible explanation is that on some platforms the semaphores need to be allocated contiguously in a single semaphore set. That is why my book says, "If the semaphore clients are not always shut down and started up in strict sequence, then an extra allowance at least equal to the largest single requirement is recommended."

I was under the impression that you need at least one semaphore per process. On one of our boxes, we are short of semaphores. We took down one 7.3.4 database with 200 processes, and two 8.1.5 databases with 100 processes each and temporarily bumped up the number of processes in the QA database by 400. At that point, we were unable to start up the QA database due to a shortage of semaphores.

Why is it that we could distribute those 400 processes among four databases and they would all come up at once but when I consolidated them in a single database, it would not come up?

I figure it must be a difference in the init.ora file of the QA database. What things might affect the ratio between processes and semaphores.

Short of increasing the semaphores in /etc/system and bouncing the box (which is not a great option), what else might we try tweaking in the QA database to enable it to come up in the current operating system environment.