0017645: corosync with two-link knet fails after "ifdown eth1"

ID	Project	Category	View Status	Date Submitted	Last Update

0017645	CentOS-8	corosync	public	2020-08-05 14:07	2020-08-05 14:07

Reporter	Yenya
Priority	normal	Severity	minor	Reproducibility	always
Status	new	Resolution	open
Product Version	8.2.2004
Target Version		Fixed in Version

Summary	0017645: corosync with two-link knet fails after "ifdown eth1"
Description	Originally reported upstream here: https://github.com/corosync/corosync/issues/569 With two-link two-node cluster, corosync correctly survives blocking one of the links by dropping all traffic using iptables, but fails when the interface is brought down by "ifdown eth1". Affected package: libknet1-1.10-4.el8
Steps To Reproduce	Have two nodes conneced to the 172.26.1.x/24 network with their eth0 interfaces, and connected to the 10.0.0.x/24 network with their eth1 interfaces. Create a pcs cluster: pcs1# pcs cluster setup yklastr \ pcs1.yenya.net addr=172.26.1.147 addr=10.0.0.147 \ pcs2.yenya.net addr=172.26.1.208 addr=10.0.0.208 Wait for it to become online: pcs1# pcs cluster status Bring down one interface: pcs2# ifdown eth1 Expected behaviour: the cluster should communicate on the eth0 network. Current behaviour: corosync on pcs2 node eats 95-100 % CPU, and the cluster communication does not work (no traffic on eth0).
Additional Information	Rebuilding kronosnet-1.15-1.fc32.src.rpm from Fedora 32, installing the RPMs, and restarting corosync fixes the problem. CentOS 8 RPMs from the upstream repository linked from the github issue above also work.
Tags	No tags attached.

View Issue Details

Activities

Issue History