> There are two common methods used by programs that depend on database > schema. One is probably best described as "YOLO", i.e. don't bother > checking anything, not ideal but it is very common. The other is to have > a schema version stored in the database and refuse to run if it's not as > expected. > > There are plenty of things in ports that require schema updates from > time to time and I don't think there's a single one that stops running > services etc.
In this case, if the collector daemon were running during an update, it'll just halt when it detects the disparity. Which also seems not expected. (The CGI not running is fine, as it'd die if the CGI were mode 0 anyway.) So the choices seem to be do nothing and have the daemon just die, which is not expected; or to have some sort of stop/start facility, which is also not expected. Upon looking back at expected behaviour, the install guide stipulates rebooting after pkg_add -u anyway. So I guess it doesn't matter that the collector dies during pkg_add since we're going to reboot, no? If that's the case, I'll add the version checking and resend.
