RowsetId is used to identify a batch load file in one replica so that it is 
unique in one replica[not tablet!!!] currenlty. Same rowsetid does not mean 
same data for different replica. It has many restrictions, for example:

  *   BE could not use rowset id as a cache key in file or block cache because 
rowsetid is not unique among different replicas.
  *   BE has to rename rowsetid when clone rowset from other BEs.
  *   During offload rowset to object storage like S3/BOS.., the object will be 
shared among replicas. It must have a unique rowsetid globally.


——
yiguolei

Reply via email to