After reading some data, it is said that HBase’s data files are in HDFS system, so it is difficult to modify and delete data in essence. In HBase, modifying and deleting data is to add a new version of the data (the timestamp is the latest), the old version of the data has not changed. HBaseShaoxing Jian says so, HBase’s health is sorted by dictionary. The problem is here. If row1 has been stored in a hfile, it has been dropped for a long time. Many data have been stored now, and a lot of hfile data have been dropped. At present, if there is oneA column update for a row1. Or delete. First put it in MemStore, then MemStore reaches the threshold drop. This time it should be to generate a small hfile file. The old files of hfile should not be updated. Because Hadoop can’t be changedDocuments. That’s the time. The new hfie will have row1, and there will be row1. in the previous hfile, so how can HBase handle it? It’s not a dictionary ordered at this time.
hbase This is what HBase said.
It is said that it is wrong. Only the same hfile can guarantee sorting by dictionary.