* Tom Dillard <tom.dillard at tekelec.com> [2006-10-17 12:12]: > I was unable to find any threads discussing SMF log file management, > so if anyone has pointers to such a discussion I would appreciate it. > > Basically I am wondering why management of the /var/svc/log files is > not built-in with logadm(1M). Is it a good idea to roll them and keep > the last X copies (and optionally compress older versions, etc.)? > > Has anyone added the hooks in to do this?
I'm not sure what aspect you're concerned about specifically. Presently, logadm(1M)'s default configuration includes a rule for the per-service logs: $ grep smf /etc/logadm.conf smf_logs -C 8 -s 1m /var/svc/log/*.log This rule is pretty naive. We can be more intelligent about managing svc.startd(1M)'s own log and provide a rotate feature, as is done for fmd(1M). The per-service logs are shared by both svc.startd and the processes composing the service; it's more difficult to envision rotation without restart--although we could just require a restart on a rotation, I suppose. That would impact availability metrics in a different way... The early, or alternate, logs that go to /etc/svc/volatile are also not managed in the current scheme. It would be good to hear about logging problems folks are hitting, to determine which shortcomings are the most pressing. - Stephen -- Stephen Hahn, PhD Solaris Kernel Development, Sun Microsystems stephen.hahn at sun.com http://blogs.sun.com/sch/
