Avoid guid pool exhaustion from guid leaks by syncing it with SM (#92) If the node running pods that use ib network is restarted, those pods' GUID are deleted from UFM but are persisted in ib-kubernetes. They become unusable and might exhaust the GUID pool if it's configured to be small enough. The solution is to sync guid pool with UFM because some GUIDs might have become free to use. Signed-off-by: amaslennikov <amaslennikov@nvidia.com> Signed-off-by: amaslennikov <amaslennikov@nvidia.com>