Github user wilderrodrigues commented on a diff in the pull request:

    https://github.com/apache/cloudstack/pull/1198#discussion_r47204055
  
    --- Diff: 
engine/orchestration/src/org/apache/cloudstack/engine/orchestration/NetworkOrchestrator.java
 ---
    @@ -2558,6 +2575,62 @@ public boolean restartNetwork(Long networkId, 
Account callerAccount, User caller
             }
         }
     
    +    /* If there are redundant routers in the isolated network, we follow 
the steps to make the network working better
    +     *  (1) destroy backup router if exists
    +     *  (2) create new backup router
    +     *  (3) destroy master router (then the backup will become master)
    +     *  (4) create a new router as backup router.
    +     */
    +    private boolean restartGuestNetworkWithRedundantRouters(NetworkVO 
network, List<DomainRouterVO> routers, ReservationContext context) throws 
ResourceUnavailableException, ConcurrentOperationException, 
InsufficientCapacityException {
    +        Account caller = CallContext.current().getCallingAccount();
    +        long callerUserId = CallContext.current().getCallingUserId();
    +
    +        // check the master and backup redundant state
    +        DomainRouterVO masterRouter = null;
    +        DomainRouterVO backupRouter = null;
    +        if (routers != null && routers.size() == 1) {
    +            masterRouter = routers.get(0);
    +        } if (routers != null && routers.size() == 2) {
    +            DomainRouterVO router1 = routers.get(0);
    +            DomainRouterVO router2 = routers.get(1);
    +            if (router1.getRedundantState() == RedundantState.MASTER || 
router2.getRedundantState() == RedundantState.BACKUP) {
    +                masterRouter = router1;
    +                backupRouter = router2;
    +            } else if (router1.getRedundantState() == 
RedundantState.BACKUP || router2.getRedundantState() == RedundantState.MASTER) {
    +                masterRouter = router2;
    +                backupRouter = router1;
    +            } else { // both routers are in UNKNOWN state
    +                masterRouter = router1;
    +                backupRouter = router2;
    +            }
    +        }
    +
    +        NetworkOfferingVO offering = 
_networkOfferingDao.findByIdIncludingRemoved(network.getNetworkOfferingId());
    +        DeployDestination dest = new 
DeployDestination(_dcDao.findById(network.getDataCenterId()), null, null, null);
    +        List<Provider> providersToImplement = 
getNetworkProviders(network.getId());
    +
    +        // destroy backup router
    +        if (backupRouter != null) {
    +            _routerService.destroyRouter(backupRouter.getId(), caller, 
callerUserId);
    +        }
    +        // create new backup router
    +        implementNetworkElements(dest, context, network, offering, 
providersToImplement);
    +
    +        // destroy master router
    +        if (masterRouter != null) {
    +            try {
    +                Thread.sleep(10000); // wait 10s for the 
keepalived/conntrackd on backup router
    --- End diff --
    
    It's a blocking call, you don't need it. Also, the keepalived is started 
with the kernel module. It would be way easier to prove the behaviour with an 
automated test.
    
    If you don't feel confident with writing the test, I can do it after 
christmas/new year and change the java code as well.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

Reply via email to