[
https://issues.apache.org/jira/browse/STORM-587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14381309#comment-14381309
]
ASF GitHub Bot commented on STORM-587:
--------------------------------------
Github user caofangkun commented on a diff in the pull request:
https://github.com/apache/storm/pull/475#discussion_r27186979
--- Diff:
storm-core/src/jvm/backtype/storm/transactional/state/TransactionalState.java
---
@@ -59,7 +59,8 @@ protected TransactionalState(Map conf, String id, Map
componentConf, String subr
.get(Config.TOPOLOGY_KRYO_REGISTER));
}
String transactionalRoot =
(String)conf.get(Config.TRANSACTIONAL_ZOOKEEPER_ROOT);
- String rootDir = transactionalRoot + "/" + id + "/" + subroot;
+ String topologyName = (String)conf.get(Config.TOPOLOGY_NAME);
--- End diff --
What if topologyName is null?
Should we give these configs default values？
Something like:
```
String topologyName = Utils.parseString(conf.get(Config.TOPOLOGY_NAME),
"default-topology-name");
```
> trident transactional state in zk should be namespaced with topology id
> -----------------------------------------------------------------------
>
> Key: STORM-587
> URL: https://issues.apache.org/jira/browse/STORM-587
> Project: Apache Storm
> Issue Type: Improvement
> Reporter: Parth Brahmbhatt
> Assignee: Sriharsha Chintalapani
>
> Currently when a trident transaction spout is initialized it creates a node
> in zk under /transactional with the spout name as the node's name. This is
> pretty dangerous as any other topology can be submitted with same spout name
> and now these 2 spouts will be overwriting each other's states. I believe it
> is better to namespace this with topologyId just like all other zk entries
> under /storm.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
