Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] The dreaded RHEL7 question

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] The dreaded RHEL7 question


Chronological Thread 
  • From: Mark Feit <>
  • To: "Brownlee, Johnathan" <>, "" <>
  • Subject: Re: [perfsonar-user] The dreaded RHEL7 question
  • Date: Fri, 7 Feb 2020 17:14:36 +0000

Brownlee, Johnathan writes:

 

I know, I know. Not supported. I get it.

 

It doesn’t have to be that way.  perfSONAR doesn’t support RHEL because none of the institutions doing the bulk of the development (ESnet, GÉANT, Indiana University, Internet2 and the University of Michigan) run it there and we haven’t been asked for it by a large-enough segment of the community to put it on our already-full dance cards.  Similarly, nobody who’s running it on RHEL has come to us and said, “we run this, it works and we’d like to contribute what we’ve learned to the project.”

 

This is one of those instances where a small amount of support from the community would go a long way toward getting it done.  Specifically, we’d need two things:

 

First is expertise from an organization that runs a production deployment of perfSONAR on RHEL to codify any differences in the installation and operation procedures and continue to do so as RHEL evolves.  That information will be folded into the documentation and the Ansible playbooks currently in development.  Since the development team already does the heavy lifting to make perfSONAR work under CentOS, I don’t imagine the effort for RHEL would consume a lot of time, especially for a shop that’s already doing what’s necessary to run it there.

 

Second is machinery to participate in the project’s test bed.  This is important because we need to make sure perfSONAR continues to run properly on the distributions we support.  At a minimum, we’d need two VMs (4 cores, 4 GiB of RAM, 20 GB of disk and an outside-facing NIC; gobs of bandwidth not required) to run the staging release for betas; one runs continuously and installs upgrades, the other is rebuilt from scratch nightly.  The ideal would be eight, with two each for the production, staging, minor development and patch development versions.  As we start to work on support for CentOS 8, there will be a need for additional machines to make sure what we do there runs properly on RHEL 8 as well.  There is a Vagrant configuration in development that will deploy and operate a full or partial set, so that end of it will become minimal work, too.

 

If you (or anyone else) are interested in either or both, please drop the development team a line ().

 

--Mark

 




Archive powered by MHonArc 2.6.19.

Top of Page