Random ethX interfaces are down when a node with Ubuntu restarts
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Invalid
|
High
|
Aleksander Mogylchenko | ||
5.1.x |
Fix Released
|
High
|
Aleksander Mogylchenko | ||
6.0.x |
Fix Committed
|
High
|
Aleksander Mogylchenko | ||
6.1.x |
Invalid
|
High
|
Aleksander Mogylchenko |
Bug Description
Faced on Jenkins CI test http://
There are no 'eth1' and 'eth2' interfaces (public and management networks) on the compute 'node-1' after node reset and booted.
That caused the inaccessible RabbitMQ from the compute node and as result there is failed OSTF test "Create volume and attach it to instance".
======= dmesg: http://
[ 6.040744] init: network-interface (eth2) pre-start process (1230) killed by TERM signal
[ 6.041678] init: network-interface (eth1) pre-start process (1231) killed by TERM signal
======= ifconfig: http://
The similar bug described on the following link: http://
" If VM has several network interfaces, then several instances of open-iscsi init script executed on system boot. Several of them try to load iscsi kernel modules and start iscsi daemon simultaneously. So one finished successfully, others - failed and stop process of setting IP on their network interfaces. "
As was suggested, I've removed open-iscsi from if-up.d and if-down.d (open-iscsi is already added to the correct runlevels):
rm /etc/network/
rm /etc/network/
After this, the service open-iscsi starts a bit later but without issues:
[ 2.854663] 8021q: adding VLAN 0 to HW filter on device eth2
[ 2.857242] 8021q: adding VLAN 0 to HW filter on device eth4
[ 2.858095] 8021q: adding VLAN 0 to HW filter on device eth1
[ 2.863143] 8021q: adding VLAN 0 to HW filter on device eth0
[ 3.053185] input: ImExPS/2 Generic Explorer Mouse as /devices/
[ 3.255104] openvswitch: module verification failed: signature and/or required key missing - tainting kernel
[ 3.255496] openvswitch: Open vSwitch switching datapath 1.10.2, built Nov 5 2014 14:18:31
[ 3.314766] init: udev-fallback-
[ 123.015589] init: failsafe main process (1290) killed by TERM signal
[ 123.038598] Loading iSCSI transport class v2.0-870.
[ 123.045391] iscsi: registered transport (tcp)
[ 123.060854] iscsi: registered transport (iser)
[ 123.063824] iscsid (2043): /proc/2043/oom_adj is deprecated, please use /proc/2043/
Steps to reproduce:
1. Deploy any cluster configuration based on Ubuntu with several network interfaces. Assign different networks to different interfaces.
2. Reboot nodes.
3. Check 'dmesg' output for 'network-interface .. killed by TERM signal'
When open-iscsi starts right after network interfaces started, it can cause termination of some interfaces.
Changed in fuel: | |
assignee: | nobody → Fuel for Openstack (fuel) |
importance: | Undecided → High |
Changed in fuel: | |
status: | New → Confirmed |
assignee: | Fuel for Openstack (fuel) → Fuel Library Team (fuel-library) |
Changed in fuel: | |
assignee: | Fuel Library Team (fuel-library) → MOS Linux (mos-linux) |
status: | Confirmed → Triaged |
Changed in fuel: | |
status: | Triaged → Incomplete |
Changed in fuel: | |
status: | Incomplete → Confirmed |
tags: | added: customer-found |
seems like this issue was reproduced again https:/ /bugs.launchpad .net/fuel/ +bug/1383319