Hi Demai, see my reply inline:
On 07/15/2015 12:18 AM, Demai Ni wrote: > I am following instruction here: > > https://cwiki.apache.org/confluence/display/BIGTOP/How+to+install+Bigtop+0.8.0+Hadoop+on+CentOS+6+with+Puppet > > > to setup a 2-node clusters. But failed at very early stage: puppet > apply. Well, quite possible that I made a small mistake somewhere, but I > have no clue. > > > the error location *site.pp:21 *points to jdk property. I do have a > jdk1.7 on my system, and was hoping puppet will automatically install > 1.6 as required by bigtop. You are right that puppet will install the jdk package based on value of this property, but you need to make this package available in the repositories of nodes you are deploying bigtop on first. That is because this jdk property is expected to be configured by an admin deploying bigtop based on java available on theirs machines. It doesn't magically try to download java package from somewhere behind the scene. So in your case, when you already have java 1.7 installed on your machines, you can * either check proper name of package for java 1.6 available in repositories of your distribution and edit the property accordingly * or use java 1.7 you already have installed on your system (but this may not work as official requirement of 0.8.0 release is java 1.6) But looking at your example, you seem not to configure your site.csv file at all. You need to properly fill out all required values as described in the mentioned page, otherwise you can't expect it to work. Hope it helps a bit. -- Martin Bukatovic
