Problem with R Snippet

[zeineb]

Hello,

I have a problem with R Snippet, my table contains 2 columns and ~850000 rows.
Error is : knime error could not detach connection to R could leak objects to other workspace
I think that can be problem because volumetry of the table, I try with sample of the table and It work.

Can you help me ?

[ScottF]

Hi @zeineb -

There are a couple of things you could try:

1. Changing the Rserve receiving buffer size limit in Preferences -> KNIME -> R. It’s set to 256 MB by default, but you can set it to zero to remove the limit.

2. Update your Rserve manually, as described here: R Problems on macOS (high sierra) since R version 3.5

[zeineb]

Hello,

Thank you for this solution. But,

• I have already Rserve receiving buffer size limit=0
• I tried to update Rserve manually but that’s not work => I have this Error

trying URL ‘http://rforge.net/src/contrib/Rserve_1.8-6.tar.gz’_
Content type ‘application/x-gzip’ length 496804 bytes (485 KB)
downloaded 485 KB

In R CMD INSTALL
Warning in strptime(xx, f, tz = tz) :
_ unable to identify current timezone ‘C’:_
please set environment variable ‘TZ’
Warning in untar2(tarfile, files, list, exdir, restore_times) :
_ failed to copy ‘Rserve/src/client’ to ‘Rserve/clients’_
* installing source package ‘Rserve’ …
Warning in system(“sh ./configure.win”) : ‘sh’ not found
ERROR: configuration failed for package ‘Rserve’
* removing ‘E:/R-3.5.1/library/Rserve’
* restoring previous ‘E:/R-3.5.1/library/Rserve’

The downloaded source packages are in
_ ‘C:\Users\zeineb.aljene\AppData\Local\Temp\RtmpyUy2AB\downloaded_packages’_
Warning message:
In install.packages(“Rserve”, , “http://rforge.net/”, type = “source”) :
_ installation of package ‘Rserve’ had non-zero exit status _

NB: I work on Windows platform*

Do you have others suggestion please ?

[Mark_Ortmann]

Hi @zeineb,

sadly I’m unable to reproduce your problem.

Could you tell me which Rserve, R and Windows version you are running and send me your KNIME log?

Would it be an option for you, to share you workflow (containing only a small sample of your data - or similar fake data) with me?

Cheers,
Mark

P. S. with similar I mean same column types

[ScottF]

Hi @zeineb -

A couple of other things to try, as described here:

R Snippets / Rserve version warnings

You may also install an earlier R version instead, e.g. 3.4.3, if you don’t depend on changes in the R 3.5 versions. Is there another way to get a compiled version of Rserve 1.8.6? You could try omitting the type="source" parameter, it may work. If it fails, it will fall back to asking you whether you want to be building from source instead, though.

[zeineb]

It’s Ok with R 3.4.3.
Thanks for your help

[mlauber71]

You could indeed try to set the Timezones. I think I remember it helpd me at certain points.
Another option could be to try ro install the package via RStudio. Sometimes that also helped.

---

https://www.r-bloggers.com/doing-away-with-“unknown-timezone”-warnings/

options(tz=“Europe/Berlin”)
Sys.setenv(TZ=“Europe/Berlin”)
Sys.getenv(“TZ”)

[mauuuuu5]

I everyone, I am having the same problem:

Knime version: 3.6
R version: 3.5.1 and Updated R Serve
In Knime preferences I’ve set the buffer size limit to cero
I am attaching an image of the process

Is not running with or without the previous cache node

Thank you

Mau

[mlauber71]

I have seen such errors on servers where the same R package was used by different users at the same time (you could check that and make sure the current knime installation ist the only one using this R folder, at least for the time of workflow execution). And you could try to run your R jobs one after the other not parallel.

[lareooreal]

I produced the same error message
“Execute failed: Could not detach connection to R, could leak objects to other workspaces.”
MRO 3.5.0, 3.5.3 and ordinary R 3.5.0, cache was set to infinity, KNIME version 3.7.1

processing a String dataframe with high dimensions.
(~41 thousand rows 34 columns). A simple R snippet doing almost nothing, just passing
my df to the output failed.

I imputed missings, which worked one time but not reproducible - sorry.

I then chopped the size of the data with row filtering and set keep all in memory). This leads to various results (sometimes errors sometimes not).
Lowering the row numbers increases the chance that no error occurs.
BUT the behavior is not reproducible. Some memory leak I guess