> Might make for a useful tutorial session? It's actually really easy to run > lots of ROACH boards on a network.
I think this would be great. I'll put it in the program somewhere! John > > Of the people attending from SKA-SA, I don't think any of us are planning > to talk about this in our main presentations. But I'd be happy to give a > little demonstration on the side for those who'd like to hear about "ROACH > best practices". A good venue for this might be the tutorial room where we > can get our hands dirty. > > The only documentation that I know about is here: > https://casper.berkeley.edu/wiki/Getting_Started_with_ROACH > which links to > https://casper.berkeley.edu/wiki/ROACH_NFS_guide > for NFS config. > > Once your server's setup, you can add additional ROACH boards very easily > by changing just one setting on each board to select network booting (from > the factory default which is to boot off the SD card). > > Jason > > > > On 19 Jun 2012, at 15:39, Gary, Dale wrote: > >> Hi All, >> >> We are working on a correlator design for a 16-antenna, dual pol system >> based on eight ROACH-2 boards. I am confident we can make the digital >> hardware work, but the network setup, boot process, and control of these >> 8 boards is a total mystery to me. I wonder if someone could give a >> summary talk at the GB CASPER workshop that covers these issues (even if >> it is no more than pointing out where to find the up-to-date >> documentation). That would certainly be helpful to me, and perhaps many >> others. >> >> Regards, >> Dale >> >> ________________________________ >> From: [email protected] >> [[email protected]] On Behalf Of John Ford >> [[email protected]] >> Sent: Tuesday, June 19, 2012 7:11 AM >> To: Jason Manley >> Cc: Casper Lists >> Subject: Re: [casper] resolv.conf setup in the ROACH filesystem >> >>> I'm not sure this is a bug. Each user will need to configure >>> /etc/network/interfaces, /etc/resolv.conf, /etc/hostname etc according >>> to >>> their specific network configurations. Most people with larger systems >>> run >>> an NFS root filesystem along with DHCP, so this gets configured >>> automatically on each board by their host server. >> >> Yes, but the following lines seem to throw a DNS wrench into the works. >> >>> >>>> search sensysnetworks.net >>>> nameserver 192.168.1.1 >>> >>> Is not the shipping default. Did your DHCP server hand this out by >>> chance? >>> AFAICT, sensysnetworks.net doesn't exist, so I'm not sure how this >>> happened! >> >> I'm almost certain this is the default from the etch filesystem I >> downloaded a couple of years ago. We haven't updated this in quite a >> while. I just wanted to point this out to anyone who had weird ssh and >> authentication problems. >> >> I'm glad it's already fixed in the released filesystems. >> >>> >>> IIRC, the shipping default is something like hostname roach with eth0 >>> not >>> configured. >> We boot with NFS, which configures the eth0. >> >> BTW, is there a solution for the DHCP client not renewing its lease when >> eth0 is configured during boot time? >> >> John >> >>> >>> Jason >>> >>> >>> >>> On 18 Jun 2012, at 23:20, Dan Werthimer wrote: >>> >>>> >>>> jason, mark, >>>> >>>> john asked if we could fix this in github. >>>> who should do this? >>>> >>>> dan >>>> >>>> >>>> >>>> ---------- Forwarded message ---------- >>>> From: John Ford <[email protected]> >>>> Date: Mon, Jun 18, 2012 at 2:08 PM >>>> Subject: [casper] resolv.conf setup in the ROACH filesystem >>>> To: [email protected] >>>> >>>> >>>> Hi all. In our Roach file systems, we have the following >>>> resolve.conf: >>>> >>>> Yes, Master<1019> more resolv.conf >>>> search sensysnetworks.net >>>> nameserver 192.168.1.1 >>>> >>>> This isn't really a Good Thing. It causes slowdowns and general >>>> lossage >>>> in the DNS system, since there's no 192.168.1.1 on our networks. >>>> >>>> We found that replacing the above with: >>>> >>>> Yes, Master<1002> more etch/etc/resolv.conf >>>> # Empty for NFS Roach config >>>> >>>> Yes, Master<1003> >>>> >>>> This fixes the problems with ssh delays. >>>> >>>> John >>>> >>>> >>>> >>>> >>> >>> >> >> >> >> > >
