Hadoop Installaion Data Node not Started

[ritch...@abrs.com.hk]

Referent to: https://stackoverflow.com/questions/3425688/why-does-the-hadoop-incompatible-namespaceids-issue-happen

I tried many of the above and wont work (incl already sudo chown hduser:hduser -R  /usr/local/Hadoop..... (where my namenode/datanode lies within), then I go to check and find out somehow the access right for **namenode** is hduser:hduser, and for **datanode** is root:root...., 

I can generate datanode after I manually change the owner, this could be due to the process executed is on behalf of hduser, then hduser's process wont have right to update folders owned by root, which is the fundamental cause. 

**Note:** This may happen as sometimes when you use SUDO to execute (at root authority) and later inconsistently use hduser (a lower authority) to execute the processes..but I have not fully teste but nord out the root cause of this problem (which happened to me once before in another hive installation.).

I would advise to use an xls to keep all those variables (including roles) - I find such approach highly useful in learning and trouble shooting complex subjects / issues.

Looking at a typical flow during installation:

   i) sudo mkdir  ......<namenode path>

  ii) sudo mkdir........<datanode path>

  iii) sudo chown hduser:hduser -R  <parent path of namenode and datanode>

a) the first two would establish the folders as owned by root (unless you also make hduser a member of root but normally you wont)

b) the last assign hduser as the owner/group to both nodes' parent path

c) during any format action to name node (hadoop name node format),  or any on the fly creation of temp folders.. then they are created by hduser

    also during all such hadoop process their right

However, hduser has no right to ANY folders created by root for whatever situations. Indeed this problem is common in many installation and configuration cases where

you need to use root to create somethings but execution is under a separate ID.