hdfs - Fault Tolerance in Hadoop -
we know in hadoop if data corrupts new replica created if node down initally works fine how namenode deals 4 replicas delete 1 node?
if newly created 1 or 1 suddenly?
in situation when 1 data node
goes down, name node
see data blocks under-replicated
, start replication other node in cluster bring replication expected level (default 3).
when corrupted node goes again blocks node seen over-replicated
4 replicas.
when block becomes over-replicated
, name node
chooses replica remove. name node
prefer not reduce number of racks host replicas, , secondly prefer remove replica data node
least amount of available disk space. may rebalancing load on cluster.
Comments
Post a Comment