Hi All,
I am planning to use zookeeper + curator as a distributed locking framework in
my project.
While assessing whether my project requirements can be met, I got these below
questions which I seek your help.
Scenario 1:
Resource hierarchy in locking:
We have notion of hierarchy in resources, such that if a resource (eg: DB
table) is locked, other requests for that object and all sub-ordinate
resources(eg: rows) should be blocked. Similarly, if sub-ordinates of a
resource is locked then the lock request for the resource should block on that.
Is there a way to configure curator to serve the above need?
Scenario 2:
Actor based reentrancy:
User U1 logged into the system from node N1. During the system
operations he acquires locks on the shared resources. If he requests the same
resource again, it will be granted by curator because the reentrancy check is
based on UUID of the thread. If the same user U1 login to the system again from
node N2, his requests will be queued because UUIDs won't match. My requirement
is that if any user logins simultaneously from different nodes(multiple jvms),
all his requests has to be dealt as if he logged in from a single node. Can
this be done by replacing UUID with UserID( in general any actor/application
based ID) in curator? If not please help by noting the correct method.
Thanks,
Naresh
