[Starlingx-discuss] [STX 1.0] Seek for alternative way to install compute node
Hi community: Since we encounter some difficult problems using PXE to boot a compute node,it just get a pxeboot subnet ip(169.254.202.xxx) instead of a management subnet（192.168.xx.xx）. As a result installation procedure will hold after load initrd file. So we wonder if there is another way to install a compute node with a standard mode controller, like install the compute using ISO manually and then issue configure commands to just enable compute sub-function. Any help will be appreciated!
Hi Gao Song: Your issue might be same as we met before [1]. We just change mgmt. subnet to 192.178.xx.xx to solve this issue. could you try this? [1] http://lists.starlingx.io/pipermail/starlingx-discuss/2019-May/004485.html Thanks. BR Austin Sun. From: gao.song [mailto:gaosong_1250@163.com] Sent: Friday, August 2, 2019 11:53 PM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] [STX 1.0] Seek for alternative way to install compute node Hi community: Since we encounter some difficult problems using PXE to boot a compute node,it just get a pxeboot subnet ip(169.254.202.xxx) instead of a management subnet（192.168.xx.xx）. As a result installation procedure will hold after load initrd file. So we wonder if there is another way to install a compute node with a standard mode controller, like install the compute using ISO manually and then issue configure commands to just enable compute sub-function. Any help will be appreciated!
Hi Sun: It seems similar, but there is a tiny difference but worth a shot following your suggestion. As we can see from /var/log/daemon.log,the compute node assigned an IP not belong to the mgmt subnet, while your second node got an right ip. 2019-07-31T16:12:47.000 controller-0 dnsmasq-dhcp[3416]: info DHCPDISCOVER(eno3) 6c:92:bf:2a:43:1e 2019-07-31T16:12:47.000 controller-0 dnsmasq-dhcp[3416]: info DHCPOFFER(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-dhcp[3416]: info DHCPREQUEST(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-dhcp[3416]: info DHCPACK(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-tftp[3416]: err error 0 TFTP Aborted received from 169.254.202.216 2019-07-31T16:12:48.000 controller-0 dnsmasq-tftp[3416]: info failed sending /pxeboot/pxelinux.0 to 169.254.202.216 Thanks! At 2019-08-03 09:59:29，"Sun, Austin" <austin.sun@intel.com> Wrote：
Hi Gao Song: Your issue might be same as we met before [1]. We just change mgmt. subnet to 192.178.xx.xx to solve this issue. could you try this?
[1] http://lists.starlingx.io/pipermail/starlingx-discuss/2019-May/004485.html
Thanks. BR Austin Sun.
From: gao.song [mailto:gaosong_1250@163.com] Sent: Friday, August 2, 2019 11:53 PM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] [STX 1.0] Seek for alternative way to install compute node
Hi community: Since we encounter some difficult problems using PXE to boot a compute node,it just get a pxeboot subnet ip(169.254.202.xxx) instead of a management subnet（192.168.xx.xx）. As a result installation procedure will hold after load initrd file. So we wonder if there is another way to install a compute node with a standard mode controller, like install the compute using ISO manually and then issue configure commands to just enable compute sub-function. Any help will be appreciated!
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
Just FYI. Finally, we solve the problem, but, not quite sure about the root cause. Maybe it relates to pxe in vlan, which caused some confusing problem. And we disable uefi mode of the compute node. 在 2019-08-03 11:07:59，"gao.song" <gaosong_1250@163.com> 写道： Hi Sun: It seems similar, but there is a tiny difference but worth a shot following your suggestion. As we can see from /var/log/daemon.log,the compute node assigned an IP not belong to the mgmt subnet, while your second node got an right ip. 2019-07-31T16:12:47.000 controller-0 dnsmasq-dhcp[3416]: info DHCPDISCOVER(eno3) 6c:92:bf:2a:43:1e 2019-07-31T16:12:47.000 controller-0 dnsmasq-dhcp[3416]: info DHCPOFFER(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-dhcp[3416]: info DHCPREQUEST(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-dhcp[3416]: info DHCPACK(eno3) 169.254.202.216 6c:92:bf:2a:43:1e 2019-07-31T16:12:48.000 controller-0 dnsmasq-tftp[3416]: err error 0 TFTP Aborted received from 169.254.202.216 2019-07-31T16:12:48.000 controller-0 dnsmasq-tftp[3416]: info failed sending /pxeboot/pxelinux.0 to 169.254.202.216 Thanks! At 2019-08-03 09:59:29，"Sun, Austin" <austin.sun@intel.com> Wrote：
Hi Gao Song: Your issue might be same as we met before [1]. We just change mgmt. subnet to 192.178.xx.xx to solve this issue. could you try this?
[1] http://lists.starlingx.io/pipermail/starlingx-discuss/2019-May/004485.html
Thanks. BR Austin Sun.
From: gao.song [mailto:gaosong_1250@163.com] Sent: Friday, August 2, 2019 11:53 PM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] [STX 1.0] Seek for alternative way to install compute node
Hi community: Since we encounter some difficult problems using PXE to boot a compute node,it just get a pxeboot subnet ip(169.254.202.xxx) instead of a management subnet（192.168.xx.xx）. As a result installation procedure will hold after load initrd file. So we wonder if there is another way to install a compute node with a standard mode controller, like install the compute using ISO manually and then issue configure commands to just enable compute sub-function. Any help will be appreciated!
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
participants (2)
-
gao.song
-
Sun, Austin