Fix: make maximum node_id configurable (default: 250)

[ipspace]

Fixes #2603

[jbemmel]

There's an issue here:

netlab/netsim/augment/nodes.py

Line 131 in 0a30a76

addrs.mac_eui[3] = node.id # ... using a difference in the 4th octet, not the last one #1954

[ipspace]

There's an issue here:

netlab/netsim/augment/nodes.py

Line 131 in 0a30a76

addrs.mac_eui[3] = node.id # ... using a difference in the 4th octet, not the last one #1954

Thank you. So he'd have to define the management MAC addresses manually, or should we also modify the 3rd octet if needed? Also, we're currently squatting on OUI assigned to Cisco, so maybe we should change that to locally-administered range as well?

[jbemmel]

Thank you. So he'd have to define the management MAC addresses manually, or should we also modify the 3rd octet if needed? Also, we're currently squatting on OUI assigned to Cisco, so maybe we should change that to locally-administered range as well?

Not completely solving the issue, but

addrs.mac_eui[3] = node.id // 256
addrs.mac_eui[4] = node.id % 256

would allow for 64K nodes.

We could adjust the OUI prefix as well (though orthogonal to this issue)

[ipspace]

addrs.mac_eui[3] = node.id // 256
addrs.mac_eui[4] = node.id % 256

That would change the OUI, and I know you want to have plenty of MAC address space for individual nodes. Going to locally administered MAC addresses seems to be the only sane long-term solution, but after all the crap we've seen while working on this project, I'm afraid of what might break.

Should we, for the moment, just add a slight addition to that footnote saying "if you increase MAX_NODE_ID beyond 255, you'll have to deal with management MAC addresses as well"?

[ssasso]

Side note: instead of "complex" math, you could use netaddr to handle mac addresses as well:

>>> base_mac = netaddr.EUI('02-00-00-00-00-00')
>>> id=1234
>>> node_mac = netaddr.EUI(int(base_mac)+id)
>>> node_mac
EUI('02-00-00-00-04-D2')

[ipspace]

Side note: instead of "complex" math, you could use netaddr to handle mac addresses as well:

That's what we had in the past. I'll leave it to @jbemmel to explain the sad history of this change 😒

[jbemmel]

addrs.mac_eui[3] = node.id // 256
addrs.mac_eui[4] = node.id % 256

That would change the OUI, and I know you want to have plenty of MAC address space for individual nodes.

Unless I’m mistaken, the OUI is in 0..2 (24 bits). Index 3 and 4 are available to change

We could also change 0..2 - but that is a different issue

A footnote would not help users deal with the macs - that’s a pain

[ipspace]

Unless I’m mistaken, the OUI is in 0..2 (24 bits). Index 3 and 4 are available to change

We were modifying byte #6. Then you wanted to modify byte #4. Now you claim #4 and #5 are OK. Could you please figure out what you really need (and why), otherwise we can go back to the old solution (along the lines of what @ssasso recommended) and call it a day.

Oh, and thanks for discussing things in a barely-related merged PR. That will make it so much easier to find them later.