Symptom
On a Nexus 7000 switch under specific circumstances when you migrate the peer-link from an M1 module to an M2 module you could experience packet loss.
Conditions
This occurs with VDC type change m1,m1xl m2,f2e to m1,m1xl And m1,m1xl to m1,m1xl,m2,f2e and need to meet this sequence.
Step 1 and 2. Reload N7K-A and change the vdc type on B to m1 m1xl only.
Step 3 - vPC comes up as inconsistent due to F2 module capacity mismatch Step 4 ? Change the VDC type on N7K-A to m1 m1xl only Step 5 ? Shutdown the vpc peer-link, pka, and vpc legs on B Step 7 ? Allocate all the interfaces on the M2 module on both switches Step 8 ? Change the vPC peer-link to use the M2 modules instead Step 9 ? Reload B (after saving config) Step 10 ? Once B comes up then no shut the peer-link and peer-keepalive only.
Then VQI program will be wrong on N7K-A vpc leg
01. Reload happened on N7K-A
02. MCEC ASSOC came (Po1 (Eth4/1)-MCT, Po2 ?MCEC (Eth4/2))
this time Po1,Po2 points to same eth-node (0x10-0x17) and programmed Eth4/1 VQI (0x30) in LTL table
03. Po1 (MCT) came up
04. Change VDC type m1,m1xl m2,f2e to m1,m1xl
Removed Mod-3 from vdc-2 LC mask ,did not clean Po2 eth-node
05. Po1 down
06. Po1 up
07. Po2 up
08. VDC type is changed back m1,m1xl to m1,m1xl,m2,f2e
Mod-3 is added back to VDc-2 LC mask and dump Po2 to mod-3
Because due to step-4 eth-node is not deleted ,we used same eth-node(0x10-0x17) and programmed Eth4/2 VQI i.e 0x2f
09. Now form MCT (Po1) Eth4/1 is removed & Eth3/1 is added
This is causing to reprogram the eth-node( 0x10-0x17) with Eth3/1 VQI (0x0)
So finally ended up Po2 LTL programming with VQI 0.
Work around can be reload mod -3 or delay 4th step until vPC legs come up on N7K-A
