If I start my server with one less database than the point where it fails
- it only takes up 4.4 MB of memory according to top. If I add one more
database to the server, then I get the error. I don't think that I am
out of physical memory - as there are 3GB of real memory available. And
if each Berkeley database has its own cache, it doesn't seem like any one
of them should be having memory problems.
I used to start this many databases with ldbm backends without an issue,
so I know it is somehow tied to Berkeley. I must be hitting the system
limit of some memory related resources, from having so many Berkeley
databases open at once, but I don't know what it would be - most of my
ulimit settings are set to "unlimited".
I'm running on RedHat 9.