We have a little under 10,000 devices and we support multiple store chains for example one store chain is labeled as a and the other as b.  So store numbers with routers and switches are named as such a0101rt1, a0101sw1, a0101sw2 and a0101ap1.  I have created groups call StoreA and StoreB with subgroups Routers, Switches and APs so that running jobs on all the devices of a particular type is easier.  I was hoping for a more dynamic way of grouping devices, because we have multiple engineers adding and removing devices all the time.

Hey josephenix, it sounds like you have a very reasonable need and use.  Be aware that dynamic groups do require higher CPU resources because of the need to process conditional logic on the device inventory.  I would think that having 10 conditional rules puts you at the extreme case.

I can appreciate your need to have the inventory dynamically updated as multiple users are adding/changing things.  However, you might be negatively impacting normal use with your current process.

I might suggest this...

Using the method I described earlier, connect to the NCM proxy and issue a 'list device" - this will get you a list of all devices in inventory - put this into an array.  Iterate over the array and send each into a query for the NCM proxy with command "list groups -type device -host $DEVICENAME" - this will show you the device groups the device is in - this can go into another array.  You can use your Linux grep commands to determine if it's in the right group.  Then use the 'add device to group' command we discussed earlier, if it isn't in the right group.

You could automate/cron this to run once every 2 hours or so.  This would allow you to programmatically map devices to static device groups and get the benefits of performance with that, but still allow you some flexibility in dynamic naming...

I hope that helps!