Since progress on LBaaS V2 has come to a halt due to the incubator
limbo, but mainly because development focus has been put on Octavia for
most of the team, I'd like to suggest discussing whether it is worth
pursuing a solution to the issue (at least what I see as an issue) of
drivers being responsible for the status management of entities.  

I'd rather that status management be taken care of at the plugin layer.
If something went wrong then the driver would raise an exception.  The
plugin would handle the set the status according to what exception
happened, usually ERROR.  Currently the problem with doing this is that
drivers can be asynchronous, so easily doing this is not an option.  I
attempted to do this before with the naive assumption that all drivers
not using the agent were synchronous.

I've got a possible solution for this problem, but a proof of concept
would first need to be done just to verify.  Plus, it would be tough to
do without the support of driver maintainers.  So I'd like to discuss
this in the meeting tomorrow morning after any major issues, which I
don't know of any.

OpenStack-dev mailing list

Reply via email to