Question

GO.db installation error

0

Entering edit mode

predeus • 0

@predeus-9207

Last seen 4.6 years ago

United States

Hello all,

I've been trying to install WGCNA on our cluster. It has module system, so I've chosen R-3.1.2 and installed bunch of libs locally (which all worked OK). Now, GO.db was the only one that failed to install. When I typed

source("https://bioconductor.org/biocLite.R")
biocLite("GO.db")

it returned
----- (download, installation logs here, no errors) ---------

** testing if installed package can be loaded
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error : .onLoad failed in loadNamespace() for 'GO.db', details:
  call: sqliteSendQuery(con, statement, bind.data)
  error: error in statement: disk I/O error

Seems like some sort of SQL problem, but I'm really at a loss here and would appreciate any suggestions.

All the best,

-- Alex

go sqlite • 6.8k views

ADD COMMENT • link updated 9.0 years ago by Daniel Lai • 0 • written 9.0 years ago by predeus • 0

score 2 · Answer 1 · 2015-11-17

To answer the OP's question, the problem is that one of the CRAN packages that GO.db depends on (via AnnotationDBI), namely RSQLite, changed to a new API when Bioconductore released a new version, which was in the middle of R release cycle. (WGCNA got bitten by this change as well.) You should make sure that the Bioconductor version (2.14 and 3.0 were both released for R 3.1) and the RSQlite package are compatible. My impression was that BioC 2.14 worked with RSQLite 0.4 through 0.11, and BioC 3.0 and higher works with RSQLite 1.0.

Martin Morgan · Answer 2 · 2015-11-17

0

Entering edit mode

James W. MacDonald 67k

@james-w-macdonald-5106

Last seen 2 days ago

United States

Unfortunately, we are unable to provide technical help for outdated versions. Have you tried installing the current version of R and Bioconductor? If not, I would start there.

ADD COMMENT • link 9.0 years ago James W. MacDonald 67k

0

Entering edit mode

Just hit this myself with a fresh install (from source) of R:

R version 3.3.0 (2016-05-03) -- "Supposedly Educational"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
> source('http://bioconductor.org/biocLite.R')
Bioconductor version 3.3 (BiocInstaller 1.22.2), ?biocLite for help
> biocLite()
BioC_mirror: https://bioconductor.org
Using Bioconductor 3.3 (BiocInstaller 1.22.2), R 3.3.0 (2016-05-03).
>
> biocLite('SKAT')
BioC_mirror: https://bioconductor.org
Using Bioconductor 3.3 (BiocInstaller 1.22.2), R 3.3.0 (2016-05-03).
Installing package(s) ‘SKAT’ [...]
** testing if installed package can be loaded
* DONE (SKAT)
> biocLite('GO.db')
BioC_mirror: https://bioconductor.org
Using Bioconductor 3.3 (BiocInstaller 1.22.2), R 3.3.0 (2016-05-03).
Installing package(s) ‘GO.db’
trying URL 'https://bioconductor.org/packages/3.3/data/annotation/src/contrib/GO.db_3.3.0.tar.gz'

** testing if installed package can be loaded
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error in sqliteSendQuery(con, statement, bind.data) : 
  error in statement: disk I/O error
Error : .onLoad failed in loadNamespace() for 'GO.db', details:
  call: sqliteSendQuery(con, statement, bind.data)
  error: error in statement: disk I/O error
Error: loading failed
Execution halted
ERROR: loading failed

This is with the session:

[1] RSQLite_1.0.0        DBI_0.4-1            BiocInstaller_1.22.2

ADD REPLY • link updated 8.5 years ago by Martin Morgan 25k • written 8.5 years ago by christopherlhartl • 0

0

Entering edit mode

Just to answer my question: this is a filesystem issue.

> .libPaths('/tmp')
> biocLite('GO.db')
** testing if installed package can be loaded
* DONE (GO.db)

So a workaround if you're on such a filesystem: put the source .tar.gz in a central location, and you'll need to freshly install GO.db (and whatever other .dbs you need) directly from source, into a temporary library on the local (/tmp or /scratch) filesystem.

ADD REPLY • link updated 8.5 years ago by Martin Morgan 25k • written 8.5 years ago by christopherlhartl • 0

0

Entering edit mode

Can you provide some information about what is unique about the default location to which you were trying to install these db packages?

ADD REPLY • link 8.5 years ago Martin Morgan 25k

score 0 · Answer 3 · 2015-11-17

Hi Alex,

First of all why did you choose R-3.1.2? The current release of Bioconductor (BioC 3.2) only works with the R-3.2 series, which is the most current R series (the latest in that series being R-3.2.2). By installing R-3.1.2, you make the choice to use BioC 3.1, which is not supported anymore (we only suport the current release).

Second, are you sure you're actually using R-3.1.2? You're trying to source() an https URL and my understanding is that this is only supported starting with R-3.2.

Before going any further, I would strongly recommend that you update to R-3.2.2. Then start R and install BioC 3.2 with:

source("https://bioconductor.org/biocLite.R")
biocLite()

That will only install a few Bioconductor packages but it's an important test. In particular it will install the BiocInstaller package. Once you have this package installed, you don't need to source() the biocLite.R script anymore to install packages, just load the BiocInstaller package (it contains the biocLite() function). So quit R, restart it, and do:

library(BiocInstaller)
biocLite("GO.db")

Please show us the complete output of the 2 commands above.

Thanks,

H.

score 0 · Answer 4 · 2015-12-18

Another potential reason is that the cluster you're using doesn't fully support sqlite [1]. I have personally experienced this issue on a Lustre cluster file system. If working off the cluster is not a possibility, potential solutions can be restricting installation (via the 'lib' option in biocLite) and analyses only to non-cluster disks (e.g. /tmp, /scratch, etc). More drastic solutions may be reconfigure/install the cluster with proper locking mechanisms [2].

[1] https://groups.google.com/a/udel.edu/forum/#!topic/hpc-ask/K15bYR2ranc

[2] http://comments.gmane.org/gmane.comp.file-systems.lustre.user/5724