Originally created by @activeshadow on Tue, 23 Mar 2021 03:20:10 GMT
If for some reason an experiment using the mirror app gets stopped in an inconsistent manner (ie. using minimega instead of phenix), the OVF Open Flow rule added to the cluster host the VM receiving all the mirrored data was running on will not get removed.
This has been observed creating a switching loop the next time an experiment using the mirror app is deployed.
I suggest we add a special stage to the mirror app that will clean up the cluster when run. Something like phenix-app-mirror fixup-cluster <bridge name>. This fixup stage simply needs to ensure none of the cluster hosts have an OVS Open Flow rule that includes both an in_port= and an action:output= option.
Originally created by @activeshadow on Tue, 23 Mar 2021 03:20:10 GMT
If for some reason an experiment using the
mirror
app gets stopped in an inconsistent manner (ie. using minimega instead of phenix), the OVF Open Flow rule added to the cluster host the VM receiving all the mirrored data was running on will not get removed.This has been observed creating a switching loop the next time an experiment using the
mirror
app is deployed.I suggest we add a special stage to the
mirror
app that will clean up the cluster when run. Something likephenix-app-mirror fixup-cluster <bridge name>
. This fixup stage simply needs to ensure none of the cluster hosts have an OVS Open Flow rule that includes both anin_port=
and anaction:output=
option.