Since you've brought this up, it's something I'm not sure I understand
properly. The physical machine behind your virtual server only has one or two serial ports, right? How do you get it talking to so many UPSs? My understanding at the moment is I have two choices: 1: Physically connect each UPS to one (different) server, and have that server do the monitoring and management for that UPS. 2: Physically connect each UPS to one (different) server, and have one central server do the monitoring and management for all UPSs.

Clearly it may be possible to physically connect two UPSs to a given server with two serial ports, but that's beside the point. Our UPSs for the most part don't have network support.

Greg.

Actually I don't even use the port on the hardware. I used a product by Lavalink (there are many on the market, cheapest I could find at the time) that is a serial to Ethernet converter. There is a driver that loads on Linux and provides virtual serial ports that talk to the converters. The nice thing is that the converter can be anywhere in the world. I have them all over the campus to monitor the UPS units in the network closets as well as on all the units that supply the servers. Down side is cost, about $100 each. But still cheaper than the network interfaces and really good for units that can only do serial. The multi port units give more bang for the buck. They also make a dual and quad port model. Other manufactures make 16 and 32 port units.

The catch is that server doing the monitoring needs to be the last machine down and the first up.

If you have servers => the number of ups units and all are able to run Nut you can have each ups plugged into a server and yet that server is plugged into two different UPS units. If you look at:

http://www.networkupstools.org/doc/2.0.1/big-servers.html

you will see the setup where the server can look to two different host to get the status of the UPS units that it is plugged into. If you have enough serial ports on machines that can run Nut you can do it as well, IF your run times and shutdown timing can be managed so that servers can get the shutdown nicely. The down side is that thing can get real complex if there are many servers.

Doug


_______________________________________________
Nut-upsuser mailing list
Nut-upsuser@lists.alioth.debian.org
http://lists.alioth.debian.org/mailman/listinfo/nut-upsuser

Reply via email to