Newbie request: stuck with installation error

Comments

6 comments

  • Avatar
    anon23740792

    i'm not sure what this means My usecase is to separate the execution engine from query optimization - can you provide a use case example? thanks

    0
    Comment actions Permalink
  • Avatar
    Candido Dessanti

    Hi @bala,

    Thanks for joining the forum and trying out our software.

    Generally speaking the /bin/sh: 1: nvcc: not found means that the PATH and LD_LIBRARY_PATH don't point to the right paths or they aren't exported so you shell is working but a script that's running on a new shell isn't seeing the two variables.

    Assuming you are on ubuntu, have you run the /usr/local/mapd-deps/mapd-deps.sh shell? Anyway you can try to export both variables before to run the make command (export PATH and export LD_LIBRARY_PATH)

    0
    Comment actions Permalink
  • Avatar
    Andrew Seidl

    As @candido.dessanti mentioned, the issue is that for some reason nvcc isn't being found in your PATH. It could be something as simple as logging out and back in to pick up the CUDA env vars, since it appears you were able to call nvcc in a separate terminal.

    There appears to be a second issue with the fatal: Not a git repository. It looks like you might have downloaded the .zip of the repo from Github, which is not currently supported. Could you instead use git to clone the repo with:

    git clone https://github.com/omnisci/omniscidb.git
    

    I've filed a ticket for the first issue to improve how nvcc gets found and used: https://github.com/omnisci/omniscidb/issues/553

    0
    Comment actions Permalink
  • Avatar
    balasubramanian Gurumurthy

    @candido.dessanti and Andrew

    Hey. Thanks for the faster response. We have already tried adding in the CUDA location into path using

    export PATH=/usr/local/cuda-10.1/bin\${PATH:+:\${PATH}} export LD_LIBRARY_PATH=/usr/local/cuda-10.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

    Even then, I get the same error. Maybe I miss something?

    @anon23740792

    Our usecase is to try splitting a SQL query and execute it in different DBMS - running on top of different execution devices - based on their costs, somewhat like the exchange operator but in the level of different DBMSes. Do let me know if you need anymore information.

    0
    Comment actions Permalink
  • Avatar
    anon45925161

    Are you able to run nvcc from the command line?

    0
    Comment actions Permalink
  • Avatar
    balasubramanian Gurumurthy

    yes. nvcc --version works

    $ nvcc --version nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2019 NVIDIA Corporation Built on Wed_Apr_24_19:10:27_PDT_2019 Cuda compilation tools, release 10.1, V10.1.168

    0
    Comment actions Permalink

Please sign in to leave a comment.