We have here two Z9264 switches running under DellOS10 entreprise 10.5.0.1P1. Their config is very close (exception made for the uplink which is not on the same port on both switches; 1/1/65 on one switch, 1/1/66 on the other one).
We have defined the same set of vlans on these switches.
The problem is the following: when we are sure that the vlan status is the same on both switch (=the output of show vlan is the same - modulo the name of the interface of the uplink which trunks all the vlans), we run snmpwalk to get the vlan allocation per port and it happens that the bitmask returned for the vlans allocation per port does not start at the same location on both switches.
For instance, here are the results for our vlan 100 on which all nodes are put by default:
% snmpwalk --hexOutputLength=0 -v1 -cpublic -m +Q-BRIDGE-MIB SWITCH1 22.214.171.124.126.96.36.199.188.8.131.52.4.100
Q-BRIDGE-MIB::dot1qVlanStaticUntaggedPorts.100 = Hex-STRING: 0F FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF
% snmpwalk --hexOutputLength=0 -v1 -cpublic -m +Q-BRIDGE-MIB SWITCH2 184.108.40.206.220.127.116.11.18.104.22.168.4.100
Q-BRIDGE-MIB::dot1qVlanStaticUntaggedPorts.100 = Hex-STRING: 03 FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF C0
When we capture this output, we are sure that all the data nodes are on this vlan on both switches.
We can see that 124 ports are set to 1, the difference is that on switch 1, the first bit for data node begins position 5 whereas on switch 2, it begins at position 7 (starting from 1 regarding bit position).
We checked that these positions are valid for each switch setting and unsettling Vlan for different interfaces.
The bitmask layout of the first switch seems to be the good one; the Z9264 has 32 ports groups which can be breakouted up to 4 ports by group + 2 10Gb/s ports. In our case we use 31 port groups for data (=124 nodes), and 1 port group for uplink (the first one, unused at the present time, will be breakouted into 2 40Gbs/ uplinks). So it seems logic that the first data port in our case is vlan-represented at bit position 5 (the first 4 positions being allocated for the first port group).
But when we set/unset the Vlan for the unused 10Gb/s port, we see that the first two position bits are used.
The only explanation I have at the moment is that this uplink difference matters and implies a different bitmask representation for the vlans.
I think that this is buggy because such bitmask interpretation should not depend on configuration (a port is used or not, but is always at the same position).
It may have no impact for most of use cases as configuring vlan via OS10 is working well, but in our case, we rely on SNMP information to get vlan allocation information.
Does anybody has noticed such SNMP variation between different Z9264 switches ?
In the meantime, I have fully aligned the configuration of these two switches (= the uplink is on the same interface), and the 2 bits shift between the two bitmasks is still the same (modified switch has been reloaded).
Software stack is the same, config is the same, no more idea, I will call the Dell support.