Okay. I misinterpreted our last emails regarding this. I was assuming no IB network connectivity for remote GPFS file system mounts, but there is literally no network connectivity. I had hoped to do NFS mounts over some management network just for testing. I'm not sure if there is a point to proceeding with this unless we are aloud to establish connectivity between the Petrel cluster and GSS. That was the whole point. Jack O'Connell Argonne Leadership Computing Facility (630)252-3610 -----Original Message----- From: "Allen, Benjamin S." <[email protected]> Date: Tue, 19 Aug 2014 11:18:08 -0500 To: Jack O'Connell <[email protected]> Cc: "[email protected]" <[email protected]>, "Allcock, William E." <[email protected]> Subject: Re: sudoers Hi jack, No problem. I think I went through the issues of this idea in an email back to you at some point, but Petrel and JLSE as a whole have to maintain isolation from production LCF. This is a key operating principal Bill and Mike have asked me to maintain since the inception of JLSE. This primarily means network isolation. As such, currently Petrel and GSS have no network connectivity outside of pubnet. I suggest we setup a new one-off testbed in the ISSF just for this use. This will have the added benefit of not disrupting Petrel. We will also need to have a discussion with Ed and Ryan about what networking should look like. As we're using them now, the ISSF file servers only have 1GbE uplinks, and are using Myrinet just for their own GPFS traffic. There really isn't any other options short of uplinking the Myrinet chassis itself with a 8x10GbE LAG, which we're really trying to avoid. If we really had to, we could setup the old tape movers as GPFS clients with 10GbE to Myrinet and 10GbE to the e600. We would be limited to 16 file servers (two S2A9900 couplets) in this configuration unless we forklift swapped ES5 for a core Myrinet chassis. In which case we would also loose Myrinet isolation between testbeds on the ES5 side of the room. Something Ed, Ryan, and I hoped to maintain. Let me know if you want to investigate these options further. Ben On Aug 18, 2014, at 10:42 AM, O'Connell, John F. <[email protected]> wrote:
Sorry for the delayed response to this, I've been out of the office. As I think I mentioned before we want to test the possibility of defining AFM enabled file sets and allowing Petrel users to run in the the GSS cache. At this point it is just a proof of concept test.
Jack O'Connell Argonne Leadership Computing Facility (630)252-3610
-----Original Message----- From: "Allen, Benjamin S." <[email protected]> Date: Wed, 30 Jul 2014 10:13:35 -0500 To: Jack O'Connell <[email protected]> Subject: Re: sudoers
You can. I will do this shortly. What are you planning on doing that requires sudo?
In general please let me know before you make any changes to any testbed so I can coordinate with other users. All of JLSE resources are shared and not scheduled, so we have todo manual scheduling of folks and keep an eye out for stepping on each other's toes.
Thanks,
Ben
On Jul 29, 2014, at 4:08 PM, O'Connell, John F. <[email protected]> wrote:
Ben, Can I be added to sudoers on the testbed clusters?
Jack O'Connell Argonne Leadership Computing Facility (630)252-3610