When ovn-controller signals that it is alive through its Chassis row, this change is notified to all clients connected to the SB database. In large deployments, this can create a lot of stress. There is a proposed solution upstream [0] to make use of a Private_Chassis table where clients won't get notified of these healthchecks. It can be leveraged as well by other agetns such as OSP Networking-ovn metadata agent through an external_ids command. More info in the commit message of the linked patch. [0] https://patchwork.ozlabs.org/patch/899608/
RFC patch sent upstream by Lucas: https://patchwork.ozlabs.org/patch/1240755/