Hello and thank you for PF_RING.
I have a problem with ZeroCopy that does not work with ZC interface label:
/opt/pfring/sbin/zbalance_ipc -i zc:eth4 -c 20 -n 14 -g 17 -m 1 -q 2048
29/Sep/2018 15:36:17 [zbalance_ipc.c:878] ERROR: pfring_zc_open_device
error [Socket operation on non-socket] Please check that zc:eth4 is up and
not already used
Without "zc" interface label it works fine:
/opt/pfring/sbin/zbalance_ipc -i eth4 -c 2 -n 14 -g 17 -m 1 -q 2048
02/Oct/2018 20:59:30 [zbalance_ipc.c:1034] Starting balancer with 14
consumer queues..
02/Oct/2018 20:59:30 [zbalance_ipc.c:1044] Run your application instances
as follows:
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@0
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@1
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@2
But it also causes an interrupt storm that consumes ksoftirqd for the
pinned CPU:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
108 root 20 0 0 0 0 R 100.0 0.0 6:19.82
ksoftirqd/16
490 root 20 0 1171116 3204 1864 S 24.5 0.0 0:04.03
zbalance_ipc
This should not happen. On another server with PF_RING 6.x the ZC
interface label works fine and there is no issue with ksoftirqd.
I use the standard load_driver.sh script to load the pf_ring and ixgbe
kernel modules.
What do you suggest I review ? Many thanks.
Hovsep
I have a problem with ZeroCopy that does not work with ZC interface label:
/opt/pfring/sbin/zbalance_ipc -i zc:eth4 -c 20 -n 14 -g 17 -m 1 -q 2048
29/Sep/2018 15:36:17 [zbalance_ipc.c:878] ERROR: pfring_zc_open_device
error [Socket operation on non-socket] Please check that zc:eth4 is up and
not already used
Without "zc" interface label it works fine:
/opt/pfring/sbin/zbalance_ipc -i eth4 -c 2 -n 14 -g 17 -m 1 -q 2048
02/Oct/2018 20:59:30 [zbalance_ipc.c:1034] Starting balancer with 14
consumer queues..
02/Oct/2018 20:59:30 [zbalance_ipc.c:1044] Run your application instances
as follows:
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@0
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@1
02/Oct/2018 20:59:30 [zbalance_ipc.c:1049] pfcount -i zc:2@2
But it also causes an interrupt storm that consumes ksoftirqd for the
pinned CPU:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
108 root 20 0 0 0 0 R 100.0 0.0 6:19.82
ksoftirqd/16
490 root 20 0 1171116 3204 1864 S 24.5 0.0 0:04.03
zbalance_ipc
This should not happen. On another server with PF_RING 6.x the ZC
interface label works fine and there is no issue with ksoftirqd.
I use the standard load_driver.sh script to load the pf_ring and ixgbe
kernel modules.
What do you suggest I review ? Many thanks.
Hovsep