accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ed Coleman (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-3806) Failing to create a table/namespace because it already exists should not be a warning
Date Mon, 23 Jul 2018 03:38:00 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-3806?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16552251#comment-16552251
] 

Ed Coleman commented on ACCUMULO-3806:
--------------------------------------

I'm not sure that I agree with this - I see that this has existed as an issue for a while,
but I not sure modifying {color:#000000}/src/main/java/org/apache/accumulo/master/tableOps/Utils.java{color}
is necessarily the best way to handle an issue seen during random walk testing. 

What about a production system? If I have an existing table and try to create a "new" table,
I think I rather have it fail with the exception that the table exists so that I am forced
to handle it.  With the exception, I can catch it and either proceed with the existing table,
or fail so that I do not corrupt existing data with whatever I was going to next. If it just
logs a message, it seems that I could be proceeding when I would not want to.

I need to look further into where this is occurring in random-walk test, maybe we should improve
the table naming so that it is not "common", or we could defeat the message in the test, or
it could be handled by modifying the logging configuration.

Thanks for looking at this.

> Failing to create a table/namespace because it already exists should not be a warning
> -------------------------------------------------------------------------------------
>
>                 Key: ACCUMULO-3806
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-3806
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: fate
>            Reporter: Josh Elser
>            Priority: Major
>              Labels: newbie
>             Fix For: 2.0.0
>
>         Attachments: 0001-ACCUMULO-3806-changed-checkTableDoesNotExist-in-accu.patch
>
>
> This is a really common occurrence when you're running randomwalk:
> {noformat}
> Failed to execute Repo, tid=63d0421f1b17b04a
> 	ThriftTableOperationException(tableId:null, tableName:nspc_001.ctt_000, op:CREATE, type:EXISTS,
description:null)
> 		at org.apache.accumulo.master.tableOps.Utils.checkTableDoesNotExist(Utils.java:54)
> 		at org.apache.accumulo.master.tableOps.PopulateZookeeper.call(PopulateZookeeper.java:54)
> 		at org.apache.accumulo.master.tableOps.PopulateZookeeper.call(PopulateZookeeper.java:30)
> 		at org.apache.accumulo.master.tableOps.TraceRepo.call(TraceRepo.java:57)
> 		at org.apache.accumulo.fate.Fate$TransactionRunner.run(Fate.java:72)
> 		at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> 		at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> 		at org.apache.accumulo.fate.util.LoggingRunnable.run(LoggingRunnable.java:35)
> 		at java.lang.Thread.run(Thread.java:745)
> {noformat}
> Concurrent table creations run: only one succeeds and the others fail. This is expected
and what FATE was designed to handle. We shouldn't be pushing these up to the monitor -- should
probably be a info or debug message.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message