https://pinot.apache.org/ logo
Join Slack
Powered by
# general
  • m

    Mayank

    12/12/2019, 3:04 AM
    check if the conf is correct, or if there's an issue with the code
  • e

    Elon

    12/12/2019, 3:11 AM
    Copy code
    controller.helix.cluster.name=pinot
    controller.host=pinot-controller-0
    controller.port=9000
    controller.vip.host=pinot-controller-0
    controller.vip.port=9000
    controller.data.dir=data/
    controller.zk.str=pinot-zookeeper:2181
    pinot.set.instance.id.to.hostname=true
    pinot.controller.storage.factory.class.gcs=org.apache.pinot.filesystem.GcsPinotFS
    pinot.controller.storage.factory.gcs.projectId=myProject
    pinot.controller.storage.factory.gcs.bucketName=myBucket
  • a

    Alex

    12/12/2019, 3:11 AM
    @User does it look right? controller.data.dir=data/ or should it be gcs://data
  • m

    Mayank

    12/12/2019, 3:11 AM
    Yeah dataDir seems incorrect
  • m

    Mayank

    12/12/2019, 3:12 AM
    Yep
  • a

    Alex

    12/12/2019, 3:12 AM
    cool, should it be fully qualified -> with a bucket name?
  • m

    Mayank

    12/12/2019, 3:13 AM
    IIRC it should be fully qualified
  • m

    Mayank

    12/12/2019, 3:13 AM
    Agree this is a bit non intuitive
  • a

    Alex

    12/12/2019, 3:13 AM
    Copy code
    <gcs://myBucket/data>
  • m

    Mayank

    12/12/2019, 3:14 AM
    We should fix the doc or make it more intuitive
  • a

    Alex

    12/12/2019, 3:14 AM
    I think @User wrote tests against new FS, and it was taking relative paths (wasn’t working with fully qualified). I assume something upstream removes bucket name?
  • m

    Mayank

    12/12/2019, 3:16 AM
    The code snippet that I pasted above seems to get the scheme based on datadir
  • m

    Mayank

    12/12/2019, 3:16 AM
    Unsure why the test won’t work
  • m

    Mayank

    12/12/2019, 3:16 AM
    If you find the issue let’s clean it up
  • a

    Alex

    12/12/2019, 3:45 AM
    Oks, will try with a different config
  • e

    Elon

    12/12/2019, 5:22 AM
    Looks like I need to add a scheme (i.e. "gs") and have different code paths for local files and "gs" scheme
  • e

    Elon

    12/12/2019, 11:50 AM
    Got it working, will update pull request tomorrow
  • m

    Mayank

    12/12/2019, 1:13 PM
    Thanks, this is great
  • e

    Elon

    12/12/2019, 5:02 PM
    So I was able to get a local tar file stored on the servers and I see that it created the temp schema directory on the controller (gcs bucket). Is there an endpoint where I can upload a file on gcs to the servers?
  • k

    Kishore G

    12/12/2019, 5:09 PM
    Yes. @User do you have a pointer to the docs for URI based upload?
  • j

    Jennifer Dai

    12/12/2019, 5:43 PM
    @User https://pinot.readthedocs.io/en/latest/segment_fetcher.html#push-hdfs-segment-to-pinot-controller this wiki may help; let me know if you have any questions
  • e

    Elon

    12/12/2019, 6:27 PM
    Thanks!
  • e

    Elon

    12/12/2019, 6:29 PM
    Have a quick question - is there a way to specify a "pinot fs" scheme with the /segments endpoint? i.e. if we have segments stored in gcs is there an endpoint that we can post to with the remote url? We are not using the hadoopfs, just going straight to gcs.
  • j

    Jennifer Dai

    12/12/2019, 6:31 PM
    do you have a gcs pinot fs right now?
  • j

    Jennifer Dai

    12/12/2019, 6:32 PM
    then yes, it will fetch to local based on the uri and then do metadata operations
  • j

    Jennifer Dai

    12/12/2019, 6:32 PM
    the segment fetcher is hooked up to pinot fs and can support all of them
  • j

    Jennifer Dai

    12/12/2019, 6:32 PM
    just need to add the correct configs in the server/controller such as:
  • m

    Mayank

    12/12/2019, 6:33 PM
    that's my understanding too
  • j

    Jennifer Dai

    12/12/2019, 6:33 PM
    https://pinot.readthedocs.io/en/latest/pluggable_storage.html
  • j

    Jennifer Dai

    12/12/2019, 6:33 PM
    sorry, wrong copy paste
1...105106107...160Latest