[
https://issues.apache.org/jira/browse/QUARKS-137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15237616#comment-15237616
]
Susan L. Cline commented on QUARKS-137:
---------------------------------------
The issue here is that the console needs to "guess" and which value to assign
to the oplet before the peek oplet. Per the
file:<git-repository>/incubator-quarks/target/docs/javadoc/quarks/metrics/Metrics.html#counter-quarks.topology.Topology-
"CounterOp oplets are inserted between every two graph vertices with the
following exceptions:
Oplets are only inserted upstream from a FanOut oplet.
If a chain of Peek oplets exists between oplets A and B, a Metric oplet is
inserted after the last Peek, right upstream from oplet B."
I believe this means there are no counters inserted *before* a peek oplet.
This means that in this test case the Peek oplet is getting the tuple count
from the counter oplet after it, whereas the PeriodicSource does not have a
counter oplet in front of or behind it, so the console does not have a way to
get a real tuple count.
Currently when this is the case the code in the console gets a maximum tuple
count for *all* counters and multiples it by 0.25.
I'm thinking that I could change the code so that if an oplet does not have a
source or target with a tuple count I could have it look one neighbor past it's
immediate neighbors (source and target).
Do you think that would be a good idea? I can try this and see how it looks /
if it makes sense.
> Tuple count: Source oplet flows into Peek. Source tuple count is less than
> Peek’s count
> ---------------------------------------------------------------------------------------
>
> Key: QUARKS-137
> URL: https://issues.apache.org/jira/browse/QUARKS-137
> Project: Quarks
> Issue Type: Bug
> Components: Console
> Reporter: May Wone
> Assignee: Susan L. Cline
> Priority: Minor
> Attachments: SourceToPeekTupleCount.doc
>
>
> Oplet 0 is a SupplierPeriodicSource with tuple count 98 and target oplet 1.
> Oplet 1 is a Peek with tuple count 415.
> Since peek does not change the tuples, it looks odd the source’s tuple count
> is less than the peek’s tuple count.
> Note that Peek's target is an implicit fanout oplet.
> Testcase quarks.test.svt.apps.FleetManagementAnalyticsClientApplication
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
