Sebastien Roy wrote: > Liane Praza wrote: >> Peter Memishian wrote: >>> Are you confident that 5 seconds is enough? >> Ditto on that question. I'd be tempted to give it more time. (As a >> data point, the network/physical:[nwam|default] start method timeouts >> are both 600 seconds. Anything under a minute would probably not be out >> of the realm of reasonable. I'd probably choose something between 10 >> and 30 seconds, but that's only a few seconds of consideration better >> than a wild guess.) > > As I mentioned in a previous message, in my testing, the time to come > online has ranged from between 0 and 1.x seconds. I can increase the > timeout to 30 just in case something nasty is going on with svc.startd.
It's really for if there are lots of other early services competing for startd's attention. That's clearly not the case on Solaris as-shipped/or OpenSolaris with the distros I know about, but Sun Cluster and Veritas both add a fair number of early-boot services. Combined with a slow system... > >> Also, before exiting on line 100, it'd be nice to print out a message >> about what went wrong so that a system that followed this codepath has >> something helpful in the logfile. > > How about the following? > > if [ "$state" != "online" ]; then > echo "The network/datalink-management service \c" > echo "could not be enabled." > exit $SMF_EXIT_ERR_CONFIG > fi I'd go with "did not come online" or "did not reach the online state", but I don't feel too strongly about it if you prefer your statement. liane
