[
https://issues.apache.org/jira/browse/NIFI-16433?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mark Payne updated NIFI-16433:
------------------------------
Description:
ExecuteSQL and ExecuteSQLRecord share a Fetch Size property. The default is 0,
and 0 is a legal value. A fetch size of 0 tells the JDBC driver to ignore the
hint, so a large query can pull the whole result set into memory on the NiFi
side.
The property uses a validator that allows zero and greater, and its description
says that zero means the hint is ignored.
The default should be 1000, and a configured value of 0 should be rejected.
When an existing flow is loaded, if the stored Fetch Size is the number 0,
change that stored value to 1000. Leave an Expression Language value as it is.
Update the description so it no longer says that zero means the hint is
ignored. QueryDatabaseTable has its own Fetch Size property and should not be
changed.
It is generally a bad idea to allow "buffer an infinite amount of data" but is
especially a bad idea to default to this.
was:
ExecuteSQL and ExecuteSQLRecord share a Fetch Size property. The default is 0,
and 0 is a legal value. A fetch size of 0 tells the JDBC driver to ignore the
hint, so a large query can pull the whole result set into memory on the NiFi
side.
The property uses a validator that allows zero and greater, and its description
says that zero means the hint is ignored.
The default should be 1000, and a configured value of 0 should be rejected.
When an existing flow is loaded, if the stored Fetch Size is the number 0,
change that stored value to 1000. Leave an Expression Language value as it is.
Update the description so it no longer says that zero means the hint is
ignored. QueryDatabaseTable has its own Fetch Size property and should not be
changed.
> ExecuteSQL Fetch Size of 0 loads the whole result into memory
> -------------------------------------------------------------
>
> Key: NIFI-16433
> URL: https://issues.apache.org/jira/browse/NIFI-16433
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
>
> ExecuteSQL and ExecuteSQLRecord share a Fetch Size property. The default is
> 0, and 0 is a legal value. A fetch size of 0 tells the JDBC driver to ignore
> the hint, so a large query can pull the whole result set into memory on the
> NiFi side.
> The property uses a validator that allows zero and greater, and its
> description says that zero means the hint is ignored.
> The default should be 1000, and a configured value of 0 should be rejected.
> When an existing flow is loaded, if the stored Fetch Size is the number 0,
> change that stored value to 1000. Leave an Expression Language value as it
> is. Update the description so it no longer says that zero means the hint is
> ignored. QueryDatabaseTable has its own Fetch Size property and should not be
> changed.
> It is generally a bad idea to allow "buffer an infinite amount of data" but
> is especially a bad idea to default to this.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
