Project

General

Profile

Actions

Bug #22445

open

ceph osd metadata reports wrong "back_iface"

Added by Stefan Kooman over 6 years ago. Updated about 6 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
Administration/Usability
Target version:
-
% Done:

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

ceph osd metadata reports wrong "back_iface". Example: ceph osd metadata 0

{
"id": 0,
"arch": "x86_64",
"back_addr": "[2001:7b8:80:1:0:1:2:1]:6820/1003144",
"back_iface": "ens2f0",
...

Reports "back_iface": "ens2f0" instead of "internal" interace of openvswitch port:
root@osd1:~$ ovs-vsctl show
c00c102b-d64d-4999-9ab9-864ca7e77b0e
Bridge uplink
Port uplink
Interface uplink
type: internal
Port ceph
tag: 901
Interface ceph
type: internal
Port "bond0"
trunks: [901]
Interface "ens2f1"
Interface "ens2f0"
ovs_version: "2.5.2"

osd1:~$ ip -6 a s ceph
8: ceph: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 state UNKNOWN qlen 1000
inet6 2001:7b8:80:1:0:1:2:1/64 scope global
valid_lft forever preferred_lft forever
inet6 fe80::947f:d0ff:fe57:bc3a/64 scope link
valid_lft forever preferred_lft forever

The interface ens2f0 is part of a (openvswitch) lacp port channel (just like ens2f1), but has no ip addresses associated with it:

2: ens2f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc mq master ovs-system portid 3cfdfea524c0 state UP group default qlen 1000
link/ether 3c:fd:fe:a5:24:c0 brd ff:ff:ff:ff:ff:ff
inet6 fe80::3efd:feff:fea5:24c0/64 scope link
valid_lft forever preferred_lft forever
3: ens2f1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc mq master ovs-system portid 3cfdfea524c1 state UP group default qlen 1000
link/ether 3c:fd:fe:a5:24:c1 brd ff:ff:ff:ff:ff:ff
inet6 fe80::3efd:feff:fea5:24c1/64 scope link

So, how come Ceph associate this inteface with the configured ip address?

Actions #1

Updated by John Spray over 6 years ago

  • Project changed from Ceph to RADOS
  • Category set to Administration/Usability

Hmm, this could well be the first time anyone's really tested the IPv6 path here.

Actions #2

Updated by cory gu about 6 years ago

John Spray wrote:

Hmm, this could well be the first time anyone's really tested the IPv6 path here.

https://github.com/ceph/ceph/pull/20128

Actions

Also available in: Atom PDF