https://pinot.apache.org/ logo
Join Slack
Powered by
# pinot-s3
  • k
    What do you mean fail over purpose
  • m
    At the moment I am having only single disk per pino in an AZ. If that disk fails I will lose all segments for that pino node
  • m
    I assume the normal solution will be to add disk in multiple AZs and replicate segments. However I am also thinking on the possibility of having segment copy in S3 as deep storage.
  • k
    yes, that happens by default if you configure s3 as the deep storage
  • k
    you have only one replica
  • m
    Ahh great, so I assume sync back will happen normally during a disk failure?
  • m
    I mean happen automatically
  • k
    yes
  • m
    Trying to understand the recovery time
  • k
    its the time for k8s to launch a new container + time to pull the segments from s3
  • m
    As this approach is cost effective
  • k
    you can use replication factor=2
  • m
    That means 2 copies on same disk
  • k
    then there will be two copies within the same AZ
  • k
    across multiple nodes
  • m
    ok great , I think I have to do that. Will try with deep storage first and see the recovery time meets RPO
  • k
    note all of these can be done on demand
  • k
    just change the replication factor in table config to 2 and invoke rebalance
  • k
    btw, we should have this conversation in troubleshooting so others can also benefit
  • m
    Sure Thanks for the details Kishore, You been very helpful as always..!!
  • a
    Hello guys, Good noon
  • k
    hello
  • a
    Hi Kishore, Hope you are doing well
  • a
    I have recently started using pinot and I was trying to integrate deep storage for historical data, I am able to store and see the data for controller but it doesn’t work for server data
  • a
    I have added configurations like class, accessKey, secretKey, endpoint , protocol region and S3PinotFS as the s3, factory classes
  • a
    My controller config also has controller.data.dir = s3://<bucket>/path controller.local.temp.dir= local path
  • a
    And server has pinot.server.instance.dataDir= local path Pinot.server.instance.segmentTarDir= local path/segment
  • a
    Not sure if I have to configure bucket details as well for server
  • k
    how are you pushing the segments to pinot
  • a
    segments are being pushed via Kafka streaming