Jordan Brown (Sun) wrote: > Ed Carpenter wrote: >> Yes there is a variable timeout_seconds that will check the run time of >> legacy scripts as well as smf services, the default value is for >> timeout_seconds is 1800 seconds, you can increase the value just to test >> if that variable is what causes the timeout. > > Hmm. So that's a timeout for the entire boot-to-single-user sequence?
No. It's the timeout for the single-user service. That service has /sbin/rcS as its method. > Our patch installation job could easily take hours, and could in theory > take days. (Zones do awful things to patch install times.) Does > anybody have any suggestions on how to map our needs into SMF? Is it > possible to dynamically stop the timer? Is there perhaps a way to say > "no explicit timeout, but if I don't tell you I'm healthy every 30 > minutes then you should assume that I'm dead"? Self-monitoring services are a future direction, though currently unscheduled. But you could start by applying patches as a separate service where you could set the timeout rather than putting your scripts in rcS. (I thought you already did, but I guess not.) liane
