Adding a compute node to an 11.5 base installation

[Seth.Dickerson]

Hi,

I have recently stood up an environment and installed IS v11.5. I have also configured this environment to use LDAP authentication and PAM authentication. The environment looks like:

Host 1: Service Tier (RHEL v7, IS v11.5)
Host 2: Engine Tier (RHEL v7, IS v11.5)
Host 3: XMETA Repository (Oracle 12)

I am attempting to add a compute node to this environment to make it an MPP environment. From what I have read there are 2 ways to do this:

1. copy-orchdist <hostname>
2. cross mount NFS

Beyond this point I am unsure of what to do? All of the IBM documentation I have read is not descriptive in the slightest (or perhaps I have not found the correct documentation). Does anyone have any instructions or documentation on where I can begin adding this compute node? Currently the environment works as expected but the requirements of this project is to add a compute node to the environment. I would imagine this is a very straight forward process......

[PaulVL]

Well, you came to the right place...

There are a few things you will need to do.

1) ensure that your engine binaries are exposed to the compute node. I would do an NSF / NAS mount.

2) you need ssh keys for your job execution user ids (batch IDs and users) from the Head Node down to the Compute Node.

3) edit $DSHOME/../PXEngine/etc/remsh (you will have to make it from the example file) and use SSH (rsh is also an option but is not used in big shops anymore for security reasons).

4) modify your APT file to have a node fastname on the compute node.

5) you might want to read about LOADL_PROCESSOR_LIST and set that.

6) user ids on Head Node must have the same UID and groups on the CN. Since you are LDAP, that should not be an issue.

7) make sure your SCRATCH DISK is also installed on the compute node.

all data mounts rw from both hosts.


that should get you started.

[Seth.Dickerson]

1) ensure that your engine binaries are exposed to the compute node. I would do an NSF / NAS mount.

What do you mean by "binaries"? Would this be the sharing out the /opt/IBM directory?

[Teej]

Please feel free to peruse your IIS documentation for instructions. One example link:

http://www.ibm.com/support/knowledgecen ... py_pe.html

-T.J.

[Seth.Dickerson]

thanks for the information it was most helpful. I ended up sharing out the main Engine nodes "/opt/IBM" and mounted it on the other Engine tier nodes. I tested out the compute nodes by modifying the default.apt file and jobs ran successfully.

Thanks!

[PaulVL]

Remember that you also need the project directories which are hopefully not located under /opt/IBM.

if this is a DEV box, then your users will also need access.
I like a shared mount for their HOME path.

Glad you are up and running.