You are not logged in.
Since the last update of irqbalance (to 0.56-3) I noticed that the irqbalance daemon does not start at all. this is observed on Core2Duo L9400 and P7350. Here is the debug, collected from the daemon:
[root@bluemoon ~]# irqbalance --debug
Package 0: cpu mask is 00000003 (workload 0)
Cache domain 0: cpu mask is 00000003 (workload 0)
CPU number 0 (workload 0)
CPU number 1 (workload 0)
Interrupt 0 (class timer) has workload 91
Interrupt 8 (class timer) has workload 0
Interrupt 44 (class storage) has workload 4
Interrupt 18 (class storage) has workload 0
Interrupt 1 (class legacy) has workload 1
Interrupt 47 (class legacy) has workload 0
Interrupt 43 (class legacy) has workload 0
Interrupt 42 (class legacy) has workload 0
Interrupt 41 (class legacy) has workload 0
Interrupt 40 (class legacy) has workload 0
Interrupt 22 (class legacy) has workload 0
Interrupt 20 (class legacy) has workload 0
Interrupt 19 (class legacy) has workload 0
Interrupt 17 (class legacy) has workload 0
Interrupt 16 (class legacy) has workload 0
Interrupt 12 (class legacy) has workload 0
Interrupt 9 (class legacy) has workload 0
Interrupt 46 (class other) has workload 76
Interrupt 23 (class other) has workload 0
-----------------------------------------------------------------------------
IRQ delta is 2787
Package 0: cpu mask is 00000003 (workload 2487)
Cache domain 0: cpu mask is 00000003 (workload 2124)
CPU number 0 (workload 2071)
Interrupt 18 (storage/0)
Interrupt 12 (legacy/1572)
Interrupt 20 (legacy/1)
Interrupt 42 (legacy/0)
Interrupt 40 (legacy/0)
Interrupt 19 (legacy/0)
Interrupt 16 (legacy/0)
CPU number 1 (workload 544)
Interrupt 44 (storage/3)
Interrupt 43 (legacy/40)
Interrupt 1 (legacy/2)
Interrupt 47 (legacy/0)
Interrupt 41 (legacy/0)
Interrupt 22 (legacy/0)
Interrupt 17 (legacy/0)
Interrupt 9 (legacy/0)
Interrupt 46 (other/361)
Interrupt 23 (other/0)
[root@bluemoon ~]#Is this behavior correct? It is certainly annoying, because on shutdown the FAIL message is generated due to the absence of a PID.
Arch Linux is more than just GNU/Linux -- it's an adventure
pkill -9 systemd
Offline
I'm just giving a quick hint of what I found about irqbalance when I searched for it a couple of days ago:
http://www.linuxquestions.org/questions … re-588979/
As far as I understand, there is no point in using irqbalance on a single-socket system.
Offline
Thanks for the reply. Following up on your links, I found this: http://www.anchor.com.au/blog/2009/04/a … ncing-act/. Apparently, irqbalance is hardcoded to run with --oneshot on a system with cache shared between cores. These include all intel chips prior to core i{5,7}. But not AMD CPUs, which always used separate cache per core.
On top of that, in the latest irqbalance package there was a bugfix, related to the number of CPU detection (before pkgrel=3 the daemon incorrectly counted the number of cores/cpus).
I guess, the right approach would be setting ONESHOT=1 in /etc/conf.d/irqbalance and correcting the comment there.
Arch Linux is more than just GNU/Linux -- it's an adventure
pkill -9 systemd
Offline