[
https://issues.apache.org/jira/browse/NIFI-2826?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15530059#comment-15530059
]
Joseph Percivall commented on NIFI-2826:
----------------------------------------
Hey [~josephxsxn] a couple thoughts:
1: The input type for enqueue time should probably be changed from epoch time
to a standard date format. We try to cater to the non-technical users of NiFi
when it comes to processor configuration and I fear most non-technical users
would have no idea what epoch time is. Also it would just be annoying to
convert to.
2: The default value should be removed and instead add a check to the
"getReceiver" method to see if ENQUEUE_TIME is set. We try our best to not
change any default behavior in non-major releases and there would be a
difference between "when the property gets populated with the default value"
and "when the receiver gets instantiated". So just a check on line 197 where if
ENQUEUE_TIME is set use it and else use "Instant.now()".
3: Looking at this, the processor could use some improvements in regards to
durable restart. Currently on every start/stop the partition receiver list is
cleared. It would be nice to keep those until the configuration changes so that
the processor could pick back up right where it left off so no data is lost.
That's outside the scope of this ticket but just putting down some thoughts
(I'll create another ticket). Also another ticket to be created is to make the
"Max number of events received" configurable (line 224).
Let me know your thoughts and we can iterate on the patch. Thanks!
> GetEventHubProcessor Support Enqueue Time
> -----------------------------------------
>
> Key: NIFI-2826
> URL: https://issues.apache.org/jira/browse/NIFI-2826
> Project: Apache NiFi
> Issue Type: Improvement
> Environment: centos6, jdk1.8, nifi 1.0.0
> Reporter: Joseph Niemiec
> Assignee: Joseph Niemiec
> Priority: Minor
> Fix For: 1.1.0
>
> Attachments:
> 0001-NIFI-2826-Adding-enqueue-time-to-GetAzureEventHub-pr.patch, patch.1
>
>
> The current GetEventHubProcessor should support allowing an enqueue time like
> eventhubs to read messages from the past. This would also support reading
> messages missed during downtime. Today the behaviour is for the processor to
> start at the enqueue of when it was started by the user missing any past
> messages.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
