00:15:29 Maybe https://github.com/TritonDataCenter/pkgsrc/issues? 00:15:40 But that's mostly me guessing. 00:27:04 looks good, thanks. 00:47:53 https://github.com/TritonDataCenter/pkgsrc/issues/356 In digging into the SMF manifest, I noticed that the process is being started via /opt/local/lib/svc/method/mariadb. What reasons are there that the process isn't being directly managed via SMF? 01:10:35 arai: thats pretty normal in order to handle pre startup actions etc. 01:15:54 toasterson: services needing to be debugged before they start, or the use of a script wrapping the binary being managed? 01:16:48 the use of a script wrapper :) Services tend to wokr usually :) 01:17:01 That's what I was hoping too. 01:18:07 And no we don't need a pid technically. But some mariadb tooling might need it. The OS has contracts to reap the processes down. 01:18:49 Perhaps I'm misreading this: What pre-startup actions are being performed by this specific startup script? In reviewing the code, it's literally setting a variable for a PID file, configuring a file descriptor limit and then, depending on the first parameter to the script, starting the process, or optionally killing it based on the stored PID. 01:19:28 Right, forgot about that. 01:19:57 Asking a mariadb dev, one moment. 01:21:38 OpenIndiana also does some install tasks in it's scrpt. 01:21:49 At first glance, the cleanest solution to this problem appears to be bypassing the script and just managing the binary directly via smf, but you make a very valid point about non-illumos related tooling potentially expecting a pidfile somewhere. 01:22:03 But I also don't see why we would need PIDFILE 01:22:04 although mysqld_safe seems to need it 01:22:53 also there is the possibilty to configure properties via SMF and have the script react to that 02:22:40 The dev I spoke with doesn't know of any specific requirement of mariadb tooling requiring a pid file. 02:46:33 https://github.com/TritonDataCenter/pkgsrc/issues/346