[
https://issues.apache.org/jira/browse/CLOUDSTACK-6420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13970463#comment-13970463
]
Murali Reddy commented on CLOUDSTACK-6420:
------------------------------------------
This can be a non-trivial change. Below snip is from dev list.
==============================================================================
On 15/04/14 3:08 AM, "Alena Prokharchyk" <[email protected]>
wrote:
Murali,
I have a question to you. In NetworkOrchestrator, implementNetwork code,
when we update the state for the network, why don¹t we use state machine
for Shared networks, but instead update it directly in the DB?
if (isSharedNetworkWithServices(network)) {
network.setState(Network.State.Implementing);
} else {
stateTransitTo(network, Event.ImplementNetwork);
}
That doesn¹t seem right to me as there should be no exception when it
comes to state update - the state machine should always be used once its
implemented for the object. If your code needs additional state
transitions for the Setup state, why can¹t you add it?
Initially network just had states and states were set explicitly. Actually
I happened to add state changes to explicitly happen through the state
machine for resource state change events for EventBus. But the problem I
had with
shared networks was, it had different life cycle (before cloudstack
supported L4-L7 services in shared networks) and shared network with L4-L7
service has life cycle similar to that of isolated network. For the
network in 'Setup' state, ImplementNetwork event shall succeed only for
shared network with services but should fail for shared network without
L4-L7 services. Similarly an implemented network in 'Implemented' state on
'ShutdownNetwork' event can go to 'Allocated' or 'Setup' state depending
on the network type (isolate/shared). I needed some sort of conditional
state transition in the state machine which is not there.
Due to time constraint I added minimal disruptive changes. Doing it
cleaner way means revamping the network states and transitions to avoid
conflicting transitions. Please do open a bug to fix it.
> Network implement: use network stateMachine instead of explicitly setting the
> state on network object
> -----------------------------------------------------------------------------------------------------
>
> Key: CLOUDSTACK-6420
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-6420
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: Network Controller
> Affects Versions: 4.4.0
> Reporter: Alena Prokharchyk
> Assignee: Murali Reddy
> Fix For: Future
>
>
> In NetworkOrchestrator, implementNetwork code, some places use state machine
> when update the network's state, some just update the network directly. Have
> to eliminate the latest, and use state machine everywhere.
> if (isSharedNetworkWithServices(network)) {
> network.setState(Network.State.Implementing);
> } else {
> stateTransitTo(network, Event.ImplementNetwork);
> }
--
This message was sent by Atlassian JIRA
(v6.2#6252)
