[
https://issues.apache.org/jira/browse/NIFI-4395?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16175387#comment-16175387
]
ASF GitHub Bot commented on NIFI-4395:
--------------------------------------
Github user ijokarumawak commented on a diff in the pull request:
https://github.com/apache/nifi/pull/2166#discussion_r140341776
--- Diff:
nifi-nar-bundles/nifi-standard-bundle/nifi-standard-processors/src/main/java/org/apache/nifi/processors/standard/GenerateTableFetch.java
---
@@ -402,16 +402,20 @@ private String getColumnStateMaxValue(String
tableName, Map<String, String> stat
return maxValue;
}
- private Integer getColumnType(String tableName, String colName) {
+ private Integer getColumnType(final ProcessContext context, String
tableName, String colName, FlowFile flowFile) {
final String fullyQualifiedStateKey = getStateKey(tableName,
colName);
Integer type = columnTypeMap.get(fullyQualifiedStateKey);
if (type == null && !isDynamicTableName) {
// If the table name is static and the fully-qualified key was
not found, try just the column name
type = columnTypeMap.get(getStateKey(null, colName));
}
+ if (type == null || columnTypeMap.size() == 0) {
+ // This means column type cache is clean after instance
reboot. We should re-cache column type
+ super.setup(context, false, flowFile);
--- End diff --
Calling `setup()` only updates `columnTypeMap`. The `type` variable will
stay being null here. Doesn't it throw ProcessException? Shouldn't we add `type
= columnTypeMap.get` after calling setup?
> GenerateTableFetch can't fetch column type by state after instance reboot
> -------------------------------------------------------------------------
>
> Key: NIFI-4395
> URL: https://issues.apache.org/jira/browse/NIFI-4395
> Project: Apache NiFi
> Issue Type: Bug
> Affects Versions: 1.3.0
> Reporter: Deon Huang
> Assignee: Deon Huang
> Fix For: 1.4.0
>
> Attachments: GenerateTableFetch_Exception.png
>
>
> The problem can easily be reproduce.
> Once GenerateTableFetch store state and encounter NiFi instance reboot.
> (Dynamic naming table by expression language)
> The exception will occur.
> The error in source code is list below.
> ```
> if (type == null) {
> // This shouldn't happen as we are populating columnTypeMap when the
> processor is scheduled or when the first maximum is observed
> throw new IllegalArgumentException("No column type found for: " +
> colName);
> }
> ```
> When this situation happened. The FlowFile will also be grab and can't
> release or observed.
> Processor can't grab existing column type from *columnTypeMap* through
> instance reboot.
> Hence will inevidible get this exception, rollback FlowFile and never success.
> QueryDatabaseTable processor will not encounter this exception due to it
> setup(context) every time,
> While GenerateTableFetch will not pass the condition and thus try to fetch
> column type from 0 length columnTypeMap.
> ---
> if (!isDynamicTableName && !isDynamicMaxValues) {
> super.setup(context);
> }
> ---
> I can take the issue if it is recognize as bug.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
