Open Stack

Hi DHCP Agent SMEs

 

We are reviewing the DHCP Agent in Icehouse, specifically for an issue where the dnsmasq host file has multiple stale or replicated (IP address) entries (as compared to the DB allocations for the subnet). We also checked the other HA dnsmasq instance and it is correctly sync’d with the DB. Right now we can only guess that the network cache of the DHCP Agent has somehow gotten out of sync with the DB due to Rabbit message loss of port_create_end or port_update_end (but there might be other reasons), which then results in a stales agent cache. A DHCP Agent restart will, of course, clear the problem.

 

Historically speaking, we are wondering what the reasoning / thoughts might have been at that time as to why a cache refresh on the DHCP Agent was not implemented. Our thoughts were that it could generate a lot of messaging traffic across Rabbit, but are there other reasons? NOTE: We can see that the code could support it under some exception handling cases, but it seems the “agent_updated” method is not triggered.

 

Does anyone have the history on this, and is there any thought on enabling such a sync? 

 

Thanks for any insights,

Randy

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160517/d347b794/attachment.html>