When an exec node reboots, the order for gridengine-exec from the Debian package causes it to try and start before nscld. The user it runs as is in LDAP, so the service fails, requiring a manual restart. This is a fixable problem and should be fixed.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
According to what I noticed in that other ticket, this normally comes up fine as is. Why exactly it didn't might be a bit of a mystery, but I did see LDAP connection errors in the logs. The weird thing is that every single hung host has this problem after reboot, and now the exact same hosts can reboot fine. This could be related to loads of screwed up puppet runs while hung, though. If so, that means this ticket is invalid.
Comment Actions
Possible fix? Create a custom unit file or dropins to add ordering config for initial boot:
Wants=nss-user-lookup.target After=nss-user-lookup.target
Comment Actions
Sounds like the idea! That should be pretty easy to try out at least in toolsbeta as well.