Ethan Quach wrote: > > > Jean McCormack wrote: >> Ethan Quach wrote: >>> Jean, Evan, >>> >>> This is not necessary a CR comment, but just a question. >>> >>> Isn't the kill call at line 420 in stop_ai_webserver() subject to >>> this same problem >>> since the ai webserver process is owned by the smf service in cases >>> where the install >>> service is started by an encompassing "svcadm enable install/server" >>> call for example? >> You are correct. Good catch. I confirmed your hypothesis by killing >> one of the webserver processes >> and the smf install service restarted with the same " Stopping >> because process received fatal signal from outside the service. ]" >> message in the log file. > > Did you kill it with SIGKILL or SIGTERM? I haven't looked at the > aiwebserver code, but > if it catches SIGTERM and dies gracefully, it shouldn't cause SMF to > restart the service. > > We appear to kill it with SIGTERM (the default) at 420. I killed it incorrectly.
Jean > > > -ethan > >> >> Jean >>> >>> >>> thanks, >>> -ethan >>> >>> >>> Jean McCormack wrote: >>>> Ethan Quach wrote: >>>>> Jean, >>>>> >>>>> Just for my information, after this fix, if an install service's >>>>> dns-sd process died >>>>> for whatever reason, what would be the steps on how to recover >>>>> from this? >>>>> Would a simple "installadm enable <svcname>" make things all right >>>>> again? >>>> Yes. Running installadm enable <svcname> would be the recover >>>> mechanism that would make >>>> things right. >>>> >>>> Jean >>>>> >>>>> >>>>> -ethan >>>>> >>>>> >>>>> Jean McCormack wrote: >>>>>> >>>>>> I need reviewers and authorization from either Dave or Sanjay to >>>>>> push for the following >>>>>> >>>>>> CR: http://defect.opensolaris.org/bz/show_bug.cgi?id=8399 >>>>>> Webrev: http://cr.opensolaris.org/~jeanm/slim_8399_release/ >>>>>> This has fixed a subset of Andre's hangs. There appear to be two >>>>>> issues causing the hangs. >>>>>> This solves the one manifested by multiple delete/disable >>>>>> services in rapid order where the >>>>>> final service is finally deleted. >>>>>> >>>>>> There is another hang caused by a delete quickly following a >>>>>> create that is caused by the >>>>>> smf service not quite getting to the online state before you >>>>>> delete the service. That is not >>>>>> addressed by this push. I would like to propose pushing the fix >>>>>> and opening a bug to address >>>>>> this second issue. This new bug would then be release noted. >>>>>> >>>>>> Jean >>>>>> _______________________________________________ >>>>>> caiman-discuss mailing list >>>>>> caiman-discuss at opensolaris.org >>>>>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >>>> >>
