IPaddr2: [findif] fails with "Command line is not complete" when interfaces have altnames or are down

[86-9]

On modern Ubuntu installations, the IPaddr2 resource agent fails during validate-all or start with an exit 5 (Not installed) error if there are unlinked/down interfaces or interfaces utilizing altname.

How to reproduce:

1. Have a network interface with administrative state UP but physical link DOWN (NO-CARRIER).
2. The interface uses altname (standard in newer systemd/udev naming schemes under Ubuntu).
3. Call IPaddr2 manually or via Pacemaker probe.

The Root Cause (from bash -x trace):
Inside the find_interface function, the script tries to parse altname attributes:

++++ ip -f inet link show dev
++++ grep altname
Command line is not complete. Try option "help"

Because $iface evaluates to empty at this specific line, the ip command syntax breaks.

Immediately afterwards, the script falls back to a global check instead of a device-specific check:

++ ip -f inet addr show
++ grep -q '[[:space:]]\(DOWN\|LOWERLAYERDOWN\)[[:space:]]'

This causes the agent to falsely report the target interface as down if any other unrelated interface on the system is currently down/unlinked, resulting in a false-positive exit 5.

Suggested Fix:

1. Quote and safeguard the device check in find_interface:

[ -n "$iface" ] && ip -f inet link show dev "$iface"

2. Restrict the status check to the actual $nic instead of a global ip addr show:

ip -f inet addr show dev "$nic" | grep -q '[[:space:]]\(DOWN\|LOWERLAYERDOWN\)[[:space:]]'

Best regards, Ingo

[oalbrigt]

This should be fixed where it detects the interface instead.

I guess that wasnt tested in an altname only scenario.

[86-9]

Here is the exact ip addr output from the node for context. It shows the altname setup along with the VLAN tagged interface (net12.4008@net12) where the script logic tripped:

ip addr show net12
18: net12: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 00:90:0b:da:4b:0d brd ff:ff:ff:ff:ff:ff
    altname enp26s0f0
    altname enx00900bda4b0d

ip addr show net12.4008
24: net12.4008@net12: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
    link/ether 00:90:0b:da:4b:0d brd ff:ff:ff:ff:ff:ff
    inet 192.168.79.149/26 brd 192.168.79.191 scope global net12.4008
       valid_lft forever preferred_lft forever

And here is the corresponding Pacemaker resource definition for completeness:

primitive ip_net12_4008 ocf:heartbeat:IPaddr2 \
    params ip="192.168.79.139" cidr_netmask="26" nic="net12.4008" \
    op monitor interval="10s"

Hopefully, this helps to pin down the exact parsing mismatch in find_interface.

Update:
The actual crash is entirely caused by the @net12 suffix from the VLAN interface (net12.4008@net12). Stripping the @ suffix completely resolves the issue:

        local real_dev=$(echo "$iface" | cut -d'@' -f1)
        local iface_altname="`$IP2UTIL -f inet link show dev $real_dev | grep "altname" | sed 's/^[[:space:]]*//; s/altname //g' | tr '\n' ' '`"

[oalbrigt]

Thanks.

[arslan9292]

Hi @86-9

There has been recent updates with the IPaddr2 agent which would avoid this issue. The commits:

• PR#2161
• PR#2127

Could you please confirm if you are using the version which has these fixes included?
Also please share the Ubuntu version, package version of iproute2.

Thank you!

[oalbrigt]

Closing. Feel free to reopen if it's still an issue after applying those PRs (or using the up-to-date agent from the main branch).