We just moved a production Perl CGI website from the global zone on a 4 core/16
cpu T2000 (with no other apps) into a 4 cpu sparse-root container on that same
T2000. The CPU load during user hours on the T2000 had reliably been running
between 1.0 and 1.5 since roughly late May; but now the container is
saturating all four cpu's it has and is clamoring for more (loads of 4.00+).
Two other changes we did were moving the website filesystems to ZFS, and
upgrading to kernel 120011-14. The website is running IPlanet 6, SP5, and is
using the inherited Perl 5.6.1 from the global zone.
The website/CGI apps themselves did not change, but the short-lived CGI scripts
are responsible for taking up the extra CPU time (as reported by dtrace):
Tracing... Hit Ctrl-C to stop.
short lived processes: 48.254 secs
total sample duration: 34.453 secs
Any idea why the CPU load of the CGI scripts would shift so drastically?
This message posted from opensolaris.org
zones-discuss mailing list