> On Mon, 4 Feb 2002, Manuel Bouyer wrote:
> > What a minute ... you're loading the tables to a 'mysqld' process, using
> > the 'mysql' command, rigth ?
> > Maybe it's just the mysql command which runs out of files descriptors ?
> > Did you try to raise the limit of the shell before starting it ?
On Mon, 4 Feb 2002, Randy Arabie wrote:
> Right. Like this:
> mysql -u someuser -p -D somedatabase < ./data.sql
> No. I didn't raise the shell limits, I rose the kernel maxfiles and
> the mysqld process proc.pid.rlimit.descriptors.soft limit.
> I'll try the shell tonight.
Summary of Last Nights activities:
1) Switched back to the GENERIC kernel.
2) Edit /etc/sysctl.conf; added a line kern.maxfiles=5000
4) Bump hard and soft proc.[mysqld_pid].rlimit.descriptors up to 2500.
5) Bump hard and soft proc.[shell_pid].rlimit.descriptors up to 2500.
6) Attempt to load the data, failure once again, same error message:
"ERROR 1017 at line 445: Can't find file: './geeklog/userinfo.frm' (errno: 23)"
7) fstat | wc -l shows 203 open files (done immediately following step 6 above.
8) fstat | grep mysqld | wc -l shows mysqld only has 64 open files.
9) The database only has 38 tables, and the datafile I'm trying to load is only
Someone suggested a few days ago that perhaps the error reported is not accurate. How
I test that?
My 'gut' says this is probably an easily resolved configuration issue, but I'm stumped on
where to go next.