Github user danny0405 commented on the issue:
https://github.com/apache/storm/pull/1874
@arunmahadevan thx for your nice review, i have checked the code
The fact is that, now for storm, we will pass in a `:scratch-topology-id`
to execute mk-assignments function for rebalance command, thus, nimbus will
just ignore the old assignment for the rebalanced topology. Finally, it will
get a new storm-id and totally new assignment on cluster.
When supervisor feels this, it will kill all the old workers for the
rebalanced topology, and launch new. In other words, the rebalanced topology
and the old one are totally two different topologies to nimbus and cluster.
Storm doesn't keep any state and just kill the old worker after the wait time
we passed in [from rebalance command].
So, the problem you suggest doesn't exist. The rebalance command is just
like a composite of kill and submit command.
Here is the code snippet:
```java
(defn do-rebalance [nimbus storm-id status storm-base]
(let [rebalance-options (:topology-action-options storm-base)]
(.update-storm! (:storm-cluster-state nimbus)
storm-id
(-> {:topology-action-options nil}
(assoc-non-nil :component->executors (:component->executors
rebalance-options))
(assoc-non-nil :num-workers (:num-workers rebalance-options))))
(mk-assignments nimbus :scratch-topology-id storm-id))
(defnk mk-assignments [nimbus :scratch-topology-id nil]
(if (is-leader nimbus :throw-exception false)
(let [conf (:conf nimbus)
storm-cluster-state (:storm-cluster-state nimbus)
^INimbus inimbus (:inimbus nimbus)
;; read all the topologies
topologies (locking (:submit-lock nimbus) (into {} (for [tid
(.active-storms storm-cluster-state)]
{tid (read-topology-details nimbus tid)})))
topologies (Topologies. topologies)
;; read all the assignments
assigned-topology-ids (.assignments storm-cluster-state nil)
existing-assignments (into {} (for [tid assigned-topology-ids]
;; for the topology which wants
rebalance (specified by the scratch-topology-id)
;; we exclude its assignment,
meaning that all the slots occupied by its assignment
;; will be treated as free slot in
the scheduler code.
(when (or (nil?
scratch-topology-id) (not= tid scratch-topology-id))
{tid (.assignment-info
storm-cluster-state tid nil)})))]`
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
