[jira] [Created] (ZEPPELIN-3461) JobProgressPoller is polling wrong interpreter if binding is changed after starting the job

2018-05-15 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-3461:


 Summary: JobProgressPoller is polling wrong interpreter if binding 
is changed after starting the job
 Key: ZEPPELIN-3461
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-3461
 Project: Zeppelin
  Issue Type: Bug
Reporter: Vipul Modi


JobProgressPoller is polling wrong interpreter if binding is changed after 
starting the job.

Steps to reproduce:
 * Start a long running spark command.
 * Change interpreter bindings of notebooks.
 * JobProgressPoller starts polling newly bound interpreter.

 

In my opinion we should either kill all running job on interpreter rebinding or 
cache Interpreter information in Job/Paragraph object.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)


[jira] [Created] (ZEPPELIN-2519) Remove watchers not present in viewport

2017-05-09 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-2519:


 Summary: Remove watchers not present in viewport
 Key: ZEPPELIN-2519
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-2519
 Project: Zeppelin
  Issue Type: Improvement
Reporter: Vipul Modi
Assignee: Vipul Modi


Currently all the watchers are enabled by default. I came across this github 
project https://github.com/wix/angular-viewport-watch to disable watchers not 
in viewport. This reduces number of watchers in notebooks with large number of 
notebooks.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Created] (ZEPPELIN-2419) Interpreter Status in Notebook

2017-04-17 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-2419:


 Summary: Interpreter Status in Notebook
 Key: ZEPPELIN-2419
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-2419
 Project: Zeppelin
  Issue Type: Task
Reporter: Vipul Modi
Assignee: Vipul Modi


It will be good to show interpreter status for interpreters bound to the 
notebook, details like interpreter is starting or running or stopped. Provide 
information on if the dependency download failed. Any other details. Details 
may vary for different interpreter group, not sure if a generic template makes 
sense.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Created] (ZEPPELIN-2412) Render action buttons at top right on mouse over or selection of paragraph.

2017-04-16 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-2412:


 Summary: Render action buttons at top right on mouse over or 
selection of paragraph.
 Key: ZEPPELIN-2412
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-2412
 Project: Zeppelin
  Issue Type: Bug
Reporter: Vipul Modi
Assignee: Vipul Modi


Currently zeppelin notebooks with large number of paragraphs add a lot of 
watchers. This lead longer digest cycles. If we show action buttons at top 
right only for selected paragraph and on mouse hover we can save a lot on 
number of watchers by using ng-if.

I would like to know others opinion before starting to work on this.

cc [~moon] 



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Created] (ZEPPELIN-2199) spark.lapply not working but SparkR:::lapply works.

2017-02-28 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-2199:


 Summary: spark.lapply not working but SparkR:::lapply works.
 Key: ZEPPELIN-2199
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-2199
 Project: Zeppelin
  Issue Type: Bug
Affects Versions: 0.7.0
Reporter: Vipul Modi


To run lapply in distributed mode via Zeppelin one needs to run SparkR:::lapply 
instead of spark.lapply.
As per spark documentation spark.lapply should work.

Steps to reproduce:

Build zeppelin using with r profiles enabled:
mvn clean install -DskipTests -Drat.skip=true -Pspark-2.0 -Phadoop-2.4 -Pyarn 
-Ppyspark -Psparkr -Pr -Pscala-2.11

Failed Paragraph
%spark.r
families <- c("gaussian", "poisson")
df <- createDataFrame(iris)
train <- function(family){ 
   model <- model <- glm(Sepal.Length ~ Sepal.Width + Species, iris, family = 
family) 
   summary(model)
}
model.summaries <- spark.lapply(families, train)

Passin paragraph:
%spark.r
families <- c("gaussian", "poisson")
df <- createDataFrame(iris)
train <- function(family){ 
   model <- model <- glm(Sepal.Length ~ Sepal.Width + Species, iris, family = 
family) 
   summary(model)
}
model.summaries <- spark.lapply(families, train)



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)


[jira] [Created] (ZEPPELIN-1364) Send details of the executed paragraph instead of complete note json.

2016-08-24 Thread Vipul Modi (JIRA)
Vipul Modi created ZEPPELIN-1364:


 Summary: Send details of the executed paragraph instead of 
complete note json.
 Key: ZEPPELIN-1364
 URL: https://issues.apache.org/jira/browse/ZEPPELIN-1364
 Project: Zeppelin
  Issue Type: Bug
Reporter: Vipul Modi


If you run a note book having lots of paragraphs and few images and svgs. Every 
paragraph run sends the complete notebook instead of sending the details of 
just the executed paragraph. We can see good performance gain if we send only 
the details of ran paragraph.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)