[
https://issues.apache.org/jira/browse/TEPHRA-228?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16020183#comment-16020183
]
ASF GitHub Bot commented on TEPHRA-228:
---------------------------------------
Github user poornachandra commented on a diff in the pull request:
https://github.com/apache/incubator-tephra/pull/42#discussion_r117843422
--- Diff:
tephra-core/src/main/java/org/apache/tephra/distributed/TransactionServiceClient.java
---
@@ -130,10 +134,11 @@ public static void doMain(boolean verbose,
Configuration conf) throws Exception
* for service discovery. Otherwise it will look for the port in the
* config and use localhost.
* @param config a configuration containing the zookeeper properties
+ * @param clientId id of the client that identifies it when it starts a
transaction
*/
@Inject
- public TransactionServiceClient(Configuration config,
- ThriftClientProvider clientProvider) {
+ public TransactionServiceClient(Configuration config,
ThriftClientProvider clientProvider,
--- End diff --
Some clients like Apache Phoenix do not use the guice modules. They
instantiate `TransactionServiceClient` directly. It will be good to have the
existing constructor too and assign the default value for the clientId in it.
> Introduce client ID that can be used to track requests per client
> -----------------------------------------------------------------
>
> Key: TEPHRA-228
> URL: https://issues.apache.org/jira/browse/TEPHRA-228
> Project: Tephra
> Issue Type: Improvement
> Components: api, core
> Reporter: Poorna Chandra
> Assignee: Gokul Gunasekaran
> Fix For: 0.12.0-incubating
>
>
> Today transaction manager does not have details of what transaction belongs
> to which client. For instance when a there are a lot of invalid transactions
> generated by a client, it is not easy to say which client generated the
> invalid transactions. Transaction manager just logs a message saying the
> transaction ID was invalidated. This makes debugging very difficult since
> there is no easy way to map the transaction ID to a program.
> Transaction APIs should allow clients to pass in client IDs for every start
> transaction call. Transaction manager can store this client ID as part of the
> transaction metadata. And when transaction manager logs messages with
> transaction ID, it can include the client ID in the message.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
