It is for the duration of the restart. The concern is what will happen to any reads which is inflight that is using the data node. Will HDFS client know the offset to start reading from where the read failed using another DN?
On Thu, Sep 18, 2014 at 1:57 PM, Ted Yu <yuzhih...@gmail.com> wrote: > How long would your datanode be out of service ? > > If the duration is long (say 1 day), you'd better decommission it first. > > Otherwise, datanode restart can be done without going through > decommissioning. > > On Thu, Sep 18, 2014 at 8:25 AM, Biju N <bijuatapa...@gmail.com> wrote: > > > Hello There, > > Can individual datanodes restarted (through service command) or do > we > > need to add it to the exclude file in NN, restart DN it, remove it from > the > > exclude file in NN as if the data node is being decommissioned and put > > back? Any feedback on this is much appreciated. > > > > Thanks, > > Biju > > >