Comment #4 on issue 326 by [email protected]: Bad interaction between
hbal, failed migrations and instance startups
http://code.google.com/p/ganeti/issues/detail?id=326
Wouldn't it be better to "annotate" the disabled state, rather than trying
to define a new one? As I was mentioning we have already too many states
for instances, and adding a new one every time we have to express something
slightly different doesn't help, and complicates all the code paths of the
instance LUs rather than just the affected one.
eg. if we add "broken" there are two problems:
1) we now assume that it's a failed migrate or failover, and the cleanup LU
can resolve it, but what if in the future we want to express one different
type of breakage?
2) all the code paths that check for the instance status (whether it's
disabled or not, whether it's admin_up or not) now need to look for the new
state and handle it
If instead we "annotated" the disabled state with "broken:instance-migrate"
then the cleanup LUs would need to be changed to know that they are allowed
to clear the disabled flag iff it's annotated that way, but all the rest of
the code would stay the same.
You're right on the fact that the cleanup job would have to be automatic,
and perhaps we'd need to change the job submission infrastructure to allow
for that. So it wouldn't be for 2.8. I guess the change above shouldn't be
in 2.8 either, actually, so we have to move this to 2.9 or 2.10. :/
Guido
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
