Hi!
I'm trying to implement a new scheduler for LVS as some
of you know already.
I'm browsing through the code for a while now, and I'm not
sure if the init/update mechanisms are very good.
In the current implementations these functions are only
used to keep track with the number of loaded modules
and to set the pointer "svc->sched_data" to a list
of valid realservers.
Did I get this right?
If I need the number of the available realservers for
my scheduling, it would be a nice idea to count them
at the init- and update-function and NOT every time
the scheduler is called, right?
The problem is, that it doesn't work. After initialising,
the "svc->sched_data" points at an empty list (size 0),
but when the scheduler is called, the list is filled
with the realservers without calling the "update".
Is this a good idea? Wouldn't it be better to call the
update after every change of the scheduling-relevant
data?
Thomas
|