I want to use Azure Disk for the Deep Store instea...
# generals
Saravanan Subburayal
04/19/2025, 9:01 AMI want to use Azure Disk for the Deep Store instead of S3 or HDFS. Are there any supported file system for storing the files locally (Azure Disk through PVC)?
x
Xiang Fu
04/19/2025, 11:15 AM
you can make AzureDisk a PVC and mount to controller then use controller
local fss
Saravanan Subburayal
04/19/2025, 1:35 PMDo we have this plugin?
org.apache.pinot.plugin.filesystem.LocalPinotFS
Saravanan Subburayal
04/19/2025, 1:37 PMI could see "org.apache.pinot.spi.filesystem.LocalPinotFS", can I use this for the ingestion in Deep Store.
x
Xiang Fu
04/19/2025, 1:39 PM
LocalPinotFS is default. If you configure the PVC and mount it on controller, then you can configure your controller deep store to this mount path
Xiang Fu
04/19/2025, 1:40 PM
Ensure you don't configure the controller vip url so that the segment download uri will be per controller basis
s
Saravanan Subburayal
04/19/2025, 1:41 PMI could see the following configuration in controller for S3.
pinot.controller.storage.factory.class.s3=org.apache.pinot.plugin.filesystem.S3PinotFS
pinot.controller.storage.factory.s3.region=us-west-2
pinot.controller.storage.factory.s3.endpoint=http://s3g-0.s3g:9878/
controller.enable.split.commit=true
pinot.controller.segment.fetcher.protocols=file,http,s3
pinot.controller.segment.fetcher.s3.class=org.apache.pinot.common.utils.fetcher.PinotFSSegmentFetcher
Do you have any reference for implementing the deepstore using localFS?
x
Xiang Fu
04/19/2025, 2:07 PM
your local fs cannot be used for split commit
Xiang Fu
04/19/2025, 2:07 PM
its only local to your controller
Xiang Fu
04/19/2025, 2:11 PM
iirc, azure disk PVC can only be mounted as RW to one pod and ReadOnly for other pods
Xiang Fu
04/19/2025, 2:11 PM
pinot deepstore with split commits etc requires the deep store to be RW for all the pods
m
Mayank
04/21/2025, 5:41 AM
Also, curious, why not use ADLS (which is supported by Pinot)?
s
Saravanan Subburayal
04/21/2025, 9:47 AMMy team member told me that it is costlier than Azure DIsks (currently we are using HDD ), let me do the pricing comparison on this again.
x
Xiang Fu
04/21/2025, 3:06 PM
ADLS hot tier is half of the price of the cheapest version of Azure disk
s
Saravanan Subburayal
04/21/2025, 4:47 PMOh ok, let me check with my team and get back to you.
4 Views