> On 14 May 2020, at 20:21, Bharath Vissapragada <bhara...@apache.org> wrote: > >> Maybe TS corruption issue some how linked with another issue that we got > - https://issues.apache.org/jira/browse/HBASE-22862 > <https://issues.apache.org/jira/browse/HBASE-22862> > > We are running into this too. Our current theory is that it is caused by > Phoenix indexes and an upgrade may fix it but I don't know if/how these two > issues are linked.
Our temporary fix for this is Phoenix table option STORE_NULLS = true. With this option Phoenix client do not generate DELETE mutations for UPSERTs with ‘NULL’ values for columns. Also this options reduce RPC and CPU load, but slightly increase size of HFiles.
