I've become bald pulling out my hair because of this problem. Time and time again I need to run an R package that needs to use RSQLite but fails because of NFS locking. Well, that's the setup on our cluster. Curiously, I don't see other people complaining about this problem but I know that a lot of bioinformatics is done on Sun Grid Engine installations, which I assume employ NFS in one way or another.
Any tips on how to alleviate this pain? Are my assumptions right (about SGE and NFS being inseparable)?
Thanks!
PK
•
written
by
polarise
This is a SQLite FAQ.
Honestly, what I would do in this situation is probably to migrate the package to a server based DBMS by patching the code, e.g. Migrate to RPostgreSQL. The SQL should be mostly compatible and run updates in a transaction.
And did I say already that it would really help to know which package you use? Giving sufficient level of details often helps solving problems, just saying...
I'm using the following package: http://bioconductor.org/packages/release/data/annotation/html/hugene10stprobeset.db.html
The current solution is to install the package on the master node (if accessible) since that will have its own disk hence no NFS lock.
Log in to answer this question.
This is definitely off topic, but I thank that on serverfault.com you will receive a good answer if you give some more details: e.g. http://serverfault.com/questions/71871/best-system-to-serve-sqlite-databases-to-webservers-over-a-network
my experience: sqlite3 doesn't work with SGE/NFS. I tried to log the steps of my NGS workflow with sqlite3, I had to switch to a java standalone sql engine (derby)
Perhaps if I stated that I was running a Bioconductor package it might draw it in... ;-)
If you could also say which package, that may make it bioinformatics
Which R/Bioconductor package are you using and when do you get this message ? (I mean are you trying to parallelize calls to a function that uses RSQLite? )