I'm developing a new stack in order to deploy my customized packages and
services.
I just started with HDFS, using HDP stack as reference. The first change
I'd like to introduce is the method used to start services. In HDP stack I
can see how the start command for services are constructed and
executed({ambari-server-path}/HDP/package/scripts/hdfs_namenode.py), but
for my understanding this logic should be delegated to hdfs namenode
service, so I'm going to use a simpler start command (service
hadoop-hdfs-namenode start), and push this logic to the service script
(/etc/init.d/hadoop-hdfs-namenode).
I also checked that using Execute class I'd miss the service output and
neither can use "service hadoop-hdfs-namenode status", so I'm thinking
about executing service operations directly with subprocess. Somethng like
this:
- File(params.exclude_file_path,
- content=Template("exclude_hosts_list.j2"),
- owner=params.hdfs_user,
- group=params.user_group
- )
-
- service(
- action="start", name="namenode", user=params.hdfs_user,
- create_pid_dir=True,
- create_log_dir=True
- )
+ p=subprocess.Popen(["service","ambari-server","status"],
stdout=subprocess.PIPE,stderr=subprocess.PIPE)
+ out,err=p.communicate()
+ rc=p.returncode
And finally add some checks with the out,err, and rc
My doubts are "am I missing some functionality or can I go with this
approach without problems?"
Regards
JC
