[
https://issues.apache.org/jira/browse/NIFI-5868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16710423#comment-16710423
]
ASF subversion and git services commented on NIFI-5868:
-------------------------------------------------------
Commit 72ea93a657cd673bfc40ef74e172f5efb950e95d in nifi's branch
refs/heads/master from [~markap14]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=72ea93a ]
NIFI-5868: Added instrumentation around ListFile such that all disk accesses
are timed and any unusually long listing times or disk access operations can be
logged. Additionally, information is logged at a debug level including
significant amounts of troubleshooting information when configured to do so
This closes #3202.
Signed-off-by: Bryan Bende <bbe...@apache.org>
> Instrument robust timing information for ListFile
> -------------------------------------------------
>
> Key: NIFI-5868
> URL: https://issues.apache.org/jira/browse/NIFI-5868
> Project: Apache NiFi
> Issue Type: Improvement
> Components: Extensions
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
> Fix For: 1.9.0
>
>
> ListFile is used in many different contexts. We often see users with a
> specific use case, though, which is to run ListFile on a Primary Node in a
> cluster, in order to obtain a file listing of an NFS-mounted share. This
> works well in most cases, but whenever problems do arise, it is very
> difficult to understand what the problem is. It would be very helpful to have
> information such as:
> * Is there a problem accessing a specific file on the NFS mount?
> * Is there a problem obtaining a listing from the NFS mount?
> * Is progress being made at all?
> * How long is a listing taking right now?
> * How long does a listing typically take?
> * Is this problem related to NiFi or to the operating system /
> infrastructure?
> It would be helpful to track information about each disk access that is
> occurring, as well as the overall listing progress and issue warnings if we
> see clear problems arise. We can do this by timing how long each disk access
> takes, what file was being accessed, and what operating was being performed.
> If we capture this data in a rolling window, we can assess the data to
> determine if the listing is now taking longer than it did previously and
> alert to this fact. Or alert if performing a specific disk operation is
> taking a long time.
> Gathering this information will likely be fairly heap intensive, so it is
> best to make the functionality optional.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
