Hi Robin,
The 'Managed' (M) flag does not appear to be set in the router advertisements.
This may be the problem if you are not administratively responsible for the router which sends the router advertisements.
Could you please provide the interface definition file(s) used by one of the CentOS 7 nodes ? (contents of /etc/sysconfig/network and /etc/sysconfig/network-scripts/ifcfg-ethX)
A small sample of logs from the DHCPv6 server would also be helpful - one, to confirm whether the client nodes are making DHCPv6 requests, and two, to determine whether the DHCPv6 server is actively trying to assign leases to the requests that it sees.
Regards,
Terry
Hi Terry,
Output attached.
Robin.
Hi Robin,
First thing would be to see what your nodes are seeing with regards to router advertisement.
tcpdump -vvvv -ttt -i eth0 icmp6 and 'ip6[40] = 134' -w racapture.pcap
(change eth0 to correct interface name)
Leave it running until you have captured at least one packet, then CTRL+C and then mail the resulting file to the list or to my personal e-mail address.
Alternatively, open it with Wireshark and verify what router advertisement flags are being set - in order for DHCPv6 to be used, the 'M' flag needs to be set.
P.S. Just seen Tim's e-mail - CentOS 7 defaults to permitting DHCPv6 responses via ip6tables, so I don't think it is that - the DUID on CentOS 7 is generated based on the MAC address and the UUID of the system BIOS, at least that is what I was able to determine when building out the lab virtual machines for my WLCG IPv6 training environment I built a year or so ago and it was indeed persistent across operating system reinstallations.
P.P.S. Once we are happy that router advertisements aren't the problem, tcpdump'ing the DHCPv6 conversation between client and server would be the next logical step.
Regards,
Terry
--Terry FroyCluster Systems Manager, Particle PhysicsQueen Mary University of LondonTel: +44 (0)207 882 6560
From: Testbed Support for GridPP member institutes <[log in to unmask]> on behalf of Robin Long <[log in to unmask]>
Sent: 18 September 2018 11:36:24
To: [log in to unmask]
Subject: Re: CentOS7 Nodes not getting v6 addressesHi Terry,
Matt is on holiday (which is why nodes have broken). Whilst I am able to do as you ask, I have no idea how to as networking is not somethign I understand, so I will need some commands to fire off.
Robin.
On 18/09/18 11:30, Terry Froy wrote:
Hi Matt,
Could you perform a packet capture between one of your DHCPv6 clients and your DHCPv6 server ? (either on the client or on the server)
It would also be useful to capture and/or provide configuration for your upstream router which provides the router advertisements to your subnet.
Quite happy to provide some sample tcpdump parameters so you can run these captures if you need them...
Regards,Terry--Terry FroyCluster Systems Manager, Particle PhysicsQueen Mary University of LondonTel: +44 (0)207 882 6560
From: Testbed Support for GridPP member institutes <[log in to unmask]> on behalf of Doidge, Matt <[log in to unmask]>
Sent: 18 September 2018 02:06:39
To: [log in to unmask]
Subject: CentOS7 Nodes not getting v6 addressesHello all,
All our Dual-Stacked CentOS 7 boxes here at Lancaster have failed to renew their DHCPv6 leases over the weekend, so we're kind of dead in the water as our "state of the art" C7, v6 enabled SE has ground to a halt. It's also hurt our perfsonar boxes, and our bdii. But any box running SL6 is having no problems with it's v6 networking. It seems OS dependent. You can ping6 from the affected machines but you can't ping6 too them (they're not binding to their v6 address). Some packet sniffing seems to show the interfaces on these machines not getting any v6 DHCP ACKnowledgements.
I've dropped a mail to our central support guys asking if they've changed anything over the last week. But to get ahead of it our end I thought I'd approach the collective. Has anyone seen anything like this before?
To give a bit more information, Lancaster is running DHCPv6 rather then SLAAC, but I've never had any luck getting a C7 boxes v6 networking to work without disabling the Network manager. In fact when NetworkManager is enabled we saw exactly the same behaviour. This setup has worked for long enough that I would have thought we'd have seen any issues by now (we've had dual-stacked perfsonars for a good few months, dual-stacked C7 DPM for a month and dual-stacked C7 BDII for a fortnight).
Any help or advice, or any idea of where to start, would be greatly appreciated.
Thanks in advance,
Matt, wondering why things always break on a week off?
########################################################################
To unsubscribe from the TB-SUPPORT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=TB-SUPPORT&A=1
To unsubscribe from the TB-SUPPORT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=TB-SUPPORT&A=1
To unsubscribe from the TB-SUPPORT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=TB-SUPPORT&A=1
To unsubscribe from the TB-SUPPORT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=TB-SUPPORT&A=1