apache/pinot

Add sanity check to the output of Realtime Provisioning Tool

开放

#8,339 创建于 2022年3月11日

 (8 条评论) (0 个反应) (0 位负责人)Java (1,234 个派生)batch import
good first issue

仓库指标

星标
 (4,937 个星标)
PR 合并指标
 (PR 指标待抓取)

描述

Currently RTP tool generates a mutable segments based on the given schema characteristics (or a given sample segment) and then linearly extrapolates the required memory, segment size, etc for different combinations of the input parameters. We've recently seen people have used 5GB segments because of the tool's suggestion. It's fine if the tool outputs 5GB because of the chosen input parameters, but it should give some warnings about potential problematic values. Here are some suggestions:

  • Change the output data structure to include an array of warnings
  • If a suggested segment size is > 500MB, add a warning message that the desired segment size should be < 500MB.
  • If number of segments is greater than a threshold (maybe 5K?), then the tools add a warning for that.
  • Sometimes tools outputs NA if the memory goes above the provided availableHostMemory parameter. Instead of NA, we should output the actual value, but add a warning that memory for some combinations goes above available host memory

贡献者指南