While upgrading is safe and works, note that 0.20.2 -> 1.0.4 is not backwards compatible. That is, existing 0.20.2 clients may fail to work with 1.0.4 services, if left un-upgraded as well.
Regarding specifying a great length of delay in block reports, it may affect a few operations that depend on most current block info at the NameNode, but I can't imagine a very critical issue being caused as a result of that. You should be OK (afaik) but certainly either apply HDFS-2379, or upgrade, soon. On Wed, Apr 10, 2013 at 1:55 PM, Xibin Liu <xibin.liu...@gmail.com> wrote: > Is upgrade from 0.20.2 to 1.0.4 safe? I have nearly 100T data and the data > are very important. > > > 2013/4/10 Harsh J <ha...@cloudera.com> > >> Xibin, >> >> I'd recommend upgrading. This was addressed quite a while ago with >> HDFS-2379. >> >> >> On Wed, Apr 10, 2013 at 1:35 PM, Xibin Liu <xibin.liu...@gmail.com>wrote: >> >>> hi all, >>> My data nodes have very large disks(20T+), and when data node reports >>> block info, the hdfs will not be able to read or write. Can I disable this >>> feature or give it a very long interval without risk? My hadoop version is >>> 0.20.2. >>> >>> -- >>> Thanks & Regards >>> Benjamin Liu >>> >>> >> >> >> -- >> Harsh J >> > > > > -- > Thanks & Regards > Benjamin Liu > > -- Harsh J