...
Impact of sticky bit and auto-recovery feature in vpc role decision should be documented for NX-OS platforms.
VPC secondary switch isolated during upgrade or for any other reason, when attached back into vPC domain and after auto-recovery timer expires may cause outage. Outage occurs as VPC roles change and vPC secondary switch takes over as Operational Primary due to sticky bit set, this leads to vPC legs to go down on Primary switch which changes to vPC operational secondary.
Disable auto-recovery prior upgrade/isolation on both vPC peers. Please note auto-recovery is by default enabled since NX-OS 6.2.x for N7k and 7.x for N5k/6k switches. Please use this command to check for sticky bit and auto-recovery status: show system internal vpcm info global | i Sticky|Reload If sticky bit is set (True) on the isolated switch, re-configuring same VPC role priority on this switch under vPC domain potentially unsets (False) the sticky bit. Only once sticky bit is false attach the isolated switch back in vPC.
NX-OS(config)# show system internal vpcm info global | i Sticky|Reload Sticky Master: TRUE >>>>> Sticky bit is set Reload timer started: FALSE Reload restore configured: TRUE, timer :240 >>> Auto-recovery enabled NX-OS(config)# show vpc role vPC Role status ---------------------------------------------------- vPC role : secondary, operational primary Dual Active Detection Status : 0 vPC system-mac : vPC system-priority : 32667 vPC local system-mac : vPC local role-priority : 200 NX-OS(config)# vpc domain 1 NX-OS(config-vpc-domain)# role priority 200 Warning: !!:: vPCs will be flapped on current primary vPC switch while attempting role change ::!! Note: --------:: Change will take effect after user has re-initd the vPC peer-link ::-------- NX-OS(config-vpc-domain)# exit NX-OS(config)# show system internal vpcm info global | i Sticky|Reload Sticky Master: FALSE >>>>>>>>>>>>>>>>> Set to False, good now to attach in vPC Reload timer started: FALSE Reload restore configured: TRUE, timer :240