Dave Williams
2016-Nov-18 10:25 UTC
[Nut-upsuser] Can this be done with NUT? (ordered shutdown, revisited)
Reference the repeater mode of the dummy-ups driver used in the way described by Steffen there is a characteristic which needs to be defined better. I could not find any notes on what the driver does in terms of shutdown. If as a result of LB you issue a upsdrvctl shutdown from a secondary server acting as a master for the repeated driver (which is pretending to run out of battery early) you DONT want it forwarding this command back to the real driver and turning that off. I looked at the driver src and it fortunately appears to be safe - but only because upsdrv_shutdown is not supported when called from the init.d scripts during poweroff. If that was ever to change then the result would depend on what was actually implemented. I suspect there might be instcmds that might be risky to forward too when the driver is used in this mode. These are OK at present as forwarding is marked as FIXME but there is an implication this will not always be the case. I am arranging a phased shutdown and currently have a topology with Server conntected to real UPS: MODE=netserver MONITOR=realups at localhost 1 upsmon secret master MINSUPPLIES=1 Server connected to repeated UPS: MODE=netserver MONITOR repeated at localhost 0 upsmon secret master MONITOR realups at server 1 upsmon secret slave MINSUPPLIES=1 Clients connect to either the primary or secondary server as netclients as per their status (critical or not). The 0 above in the MONITOR line is to provide protection for this situation and fortunaely I want the secondary server to stay up until the bitter end anyway. If that wasnt the case and I wanted to shut down this server early then I would remove the second MONITOR line and have the usual 1 setting. In that case the above concerns apply. Is there any better way of arranging the configuration? Incidentally as part of this project I have been developing a series of python scripts that can be called to iterate over a VMWare ESXi/VCenter and instigate a controlled shutdown of the VM's being manage and then shut down the ESXi itself. This uses VMWare's SDK and avoids having to add something to VMWare itself. If this is of interest I am happy to publish it here once complete for the convenience of other NUT users.
Possibly Parallel Threads
- Can this be done with NUT? (ordered shutdown, revisited)
- Can this be done with NUT? (ordered shutdown, revisited)
- debian 8 "jessie" nut 2.7.2 slaves not shutting down
- New improved version of nut-report script
- KIN-1500AP RM + Windows 2008 64-bit Standard R2 SP1 (nut 2.6.1)