I have a cluster with Ominpath nodes and CentOS 7.2.
On Sunday I upgraded the irqbalance package to irqbalance-1.0.7-6
The system is now repeatedly logging /usr/sbin/irqbalance: irq NN affinity_hint and banned cpus conflict
In /etc/sysconfig/irqbalance I have set IRQBALANCE_ARGS=--hintpolicy=exact
Looking at Redhat bugzilla, this seems to be a know issue. Has anyone seen it and is there a known workaround?
Dimitri, this is indeed an OS issue.
I have a bug opened on CentOS for it, and someone else reports the same thing on am Omnipath equipped cluster.
setting irqbalance=ignore then there are no messages logged when irqbalance is started
Ok, so root cause is identified.
In Intel® Omni-Path Performance Tuning User Guide we have the recommendations:
Setting --hintpolicy to exact is needed to work with the Receive and SDMA interrupt algorithms in the HFI1 driver.
So probably you need to wait for OS fix.