vm.exec_map_entries vs. hw.ncpu vs. ...?
David Wolfskill
david at catwhisker.org
Tue Jan 3 22:16:13 UTC 2012
A colleague at work has been researching the nature of some issues we're
seeing in a FreeBSD/amd64 release/8.2.0 environment in which amd (which
is providing services to a jail, where the "useful work" actually gets
done) is apparently failing (on occasion) to perform the work it is
supposed to do because vm.exec_map_entries is set too low.
vm.exec_map_entries default value is 16 -- see src/sys/vm/vm_init.c:91-94:
static int exec_map_entries = 16;
TUNABLE_INT("vm.exec_map_entries", &exec_map_entries);
SYSCTL_INT(_vm, OID_AUTO, exec_map_entries, CTLFLAG_RD, &exec_map_entries, 0,
"Maximum number of simultaneous execs");
We had not previously modified the setting.
The machines in question are dual-package, 6-core machines with Intel's
SMT (Simultaneous Multi-Threading) enabled, so hw.ncpu is 24.
Is it reasonable for exec_map_entries to be thus initialized,
independently of number of logical cores, kern.maxusers, amount of
memory, or anything else? Are there guidelines or suggestions for
setting it?
Apparently each entry consumes a bit over 200KB, so making it
arbitrarily large is probably not a good idea. :-}
I don't see it mentioned in tuning(7).
Thanks for any inisight.
Peace,
david
--
David H. Wolfskill david at catwhisker.org
Depriving a girl or boy of an opportunity for education is evil.
See http://www.catwhisker.org/~david/publickey.gpg for my public key.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-performance/attachments/20120103/07a6d707/attachment.pgp
More information about the freebsd-performance
mailing list