[jira] [Updated] (NIFI-1644) If unable to write to Content Repository, Process Session should automatically roll itself back

["Joseph Witt (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/NIFI-1644?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Joseph Witt updated NIFI-1644:
------------------------------
    Fix Version/s:     (was: 1.0.0)

> If unable to write to Content Repository, Process Session should automatically roll itself back
> -----------------------------------------------------------------------------------------------
>
>                 Key: NIFI-1644
>                 URL: https://issues.apache.org/jira/browse/NIFI-1644
>             Project: Apache NiFi
>          Issue Type: Bug
>          Components: Core Framework
>            Reporter: Mark Payne
>
> If we write to the Content Repository and get an IOException (for example, out of disk space), the ProcessSession catches this and then removes the temporary content claim and then throws a FlowFileAccessException. However, the entire session really should be rolled back, because the FlowFIle no longer has a valid Content Claim. An example of this is in the StandardProcessSession.write method:
> {code}
> } catch (final FlowFileAccessException ffae) {
>             resetWriteClaims(); // need to reset write claim before we can remove the claim
>             destroyContent(newClaim);
>             throw ffae;
> }
> {code}
> Processors that then catch Throwable or the general Exception and route to failure pass along an invalid FlowFile. We end up seeing the following in the logs:
> {code}
> 2016-03-17 04:21:04,742 WARN [Timer-Driven Process Thread-17] o.a.n.c.r.WriteAheadFlowFileRepository Repository Record StandardRepositoryRecord[UpdateType=CONTENTMISSING,Record=StandardFlowFileRecord[uuid=01efcc28-e28f-45ab-9373-cba8933a010c,claim=StandardContentClaim [resourceClaim=StandardResourceClaim[id=1458188464723-45209, container=pub1, section=153], offset=0, length=1017],offset=0,name=26304456091229115,size=1017]] is marked to be aborted; it will be persisted in the FlowFileRepository as a DELETE record
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)