[Beowulf] Disk less machines booted with static address fails on expired dhcp lease

[tegner at renget.se]

Hi, thanks!


Don't think it is bios related, the Dell workstations go down at the same time, while the supermicro server stay up. And these machines are booted in the same way. I have also confirmed that the time is related to the lease time I give in dhcpd.conf.


Haven't found anything in the logs, but now I start to suspect it has to do with how I let them boot, I use


ip=dhcp


would suspect it to work differently if I instead use static values here.


But then again, ip=dhcp is "easier", and why should it work on some machines, and not on others? And using infinite lease time seems to solve the issue (but a bit more "ugly").


/jon

On 22 October 2019 at 09:00:58 +02:00, Tony Brian Albers <tba at kb.dk> wrote:

> On Tue, 2019-10-22 at 08:39 +0200, <tegner at renget.se> wrote:
> 
> > Hi,
> > 
> > have been using disk less machines extensively, and just recently
> > experienced a problem connected to the dhcp lease time.
> > 
> > Initial contact is over dhcp (tied to mac address), but once the
> > machines boot over the nfs exported root file system static ip is
> > used (with the same address as initially used by dhcp).
> > 
> > This strategy has previously worked without any issues, but recently,
> > when setting up a small experimental environment using older Dell
> > desktop machines I have noted that these machines fail when the lease
> > time for the dhcp server expires.
> > 
> > This ONLY happens on the desktop machines - the server grade machines
> > I have used this far are not affected.
> > 
> > By putting infinite lease time in dhcpd.conf the problem goes away -
> > but I still would like to understand why this happens on just some of
> > the machines.
> > 
> > Regards,
> > /jon
> > 
> > _______________________________________________
> > Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin
> > Computing
> > To change your subscription (digest mode or unsubscribe) visit https:
> > //beowulf.org/cgi-bin/mailman/listinfo/beowulf
> > 
> Hejsan Tegner,
> 
> It could be something as stupid as BIOS power management shutting down
> the NIC to save electricity.
> But then again, it could be a hardware profile in the OS that doesn't
> let you keep using the static IP from the root FS.
> 
> Nothing in the logs?
> 
> /tony
> 
> 
> -- 
> Tony Albers - Systems Architect - IT Development
> Royal Danish Library, Victor Albecks Vej 1, 8000 Aarhus C, Denmark
> Tel: +45 2566 2383 - CVR/SE: 2898 8842 - EAN: 5798000792142
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://beowulf.org/pipermail/beowulf/attachments/20191022/f97672cf/attachment.html>