08-02-2018 07:05 PM
We have noticed that Cloudera Manager sets dfs.client.block.write.replace-datanode-on-failure.enable setting of HBase region servers to false by default. Can you explain why the decision was made? If I understood correctly, with the setting turned off, HBase can produce under-replicated blocks and we are concerned that it will increase the probability of data loss on some series of unfortunate events. So we decided to override the Cloudera default and turn it back on, but we'd like to know if there are some valid reasons that we are unaware of to keep it turned off.