2240-4 100% CPU on 10G CIFS transfer

hochschuleda · ‎2016-12-14

Hi,

we have an 2240-4 with a 10G Dual Mezzanin card installed. 8.2.3 7-Mode

Usually the system is just under load at night, exclusivly for 95 % snapmirror destination and 5 % CIFS backups.

Now if i disabled our normal 2x1G RJ45 VIF, and start 1 (!!!!) CIFS transfer from a fast source to the 2240 i end up having about only 50mb/s AND 100 % CPU load. If i stop the transfer, usage is back to 14% CPU.

any ideas ?

Shelf 0: DS4246 Firmware rev. IOM6E A: 0172 B: ----
Shelf 1: DS4243 Firmware rev. IOM3 A: 0212 IOM3 B: 0212

2 TB SATA Disks only.

Switch between the 2 systems is mellanox sn2700

Is it just that heavily loaded because of 1 slow transfer via 10G ???

hochschuleda · ‎2016-12-15

Without transfer or load

CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP Disk OTHER FCP iSCSI FCP kB/s iSCSI kB/s
in out read write read write age hit time ty util in out in out
11% 0 0 0 0 2 1 6908 8 0 0 0s 98% 0% - 48% 0 0 0 0 0 0 0
10% 0 0 0 0 3 1 6568 0 0 0 0s 98% 0% - 51% 0 0 0 0 0 0 0
11% 0 0 0 9 3 1 6556 0 0 0 0s 99% 0% - 45% 9 0 0 0 0 0 0
9% 0 2 0 2 2 1 7424 0 0 0 0s 98% 0% - 49% 0 0 0 0 0 0 0
11% 0 0 0 6 1 1 6960 0 0 0 0s 98% 0% - 47% 6 0 0 0 0 0 0
10% 0 0 0 9 2 1 7392 0 0 0 0s 98% 0% - 56% 9 0 0 0 0 0 0
11% 0 0 0 0 1 1 7240 0 0 0 0s 98% 0% - 47% 0 0 0 0 0 0 0
10% 0 0 0 5 1 1 7056 0 0 0 0s 98% 24% Ts 52% 5 0 0 0 0 0 0
11% 0 0 0 9 1 0 12160 6060 0 0 1 98% 42% : 54% 9 0 0 0 0 0 0
10% 0 0 0 7 2 1 7156 0 0 0 1 98% 0% - 52% 7 0 0 0 0 0 0
11% 0 1 0 2 2 1 7216 8 0 0 1 98% 0% - 52% 1 0 0 0 0 0 0
17% 0 0 0 203 3 1 6760 0 0 0 38s 98% 0% - 55% 203 0 0 0 0 0 0

1 Transfer via CIFS

98% 0 838 0 844 56176 318 71756 104708 0 0 50s 98% 78% : 64% 6 0 0 0 0 0 0
97% 0 836 0 836 54153 314 63196 86760 0 0 50s 98% 54% Hf 74% 0 0 0 0 0 0 0
97% 0 865 0 875 58172 320 44726 23797 0 0 1 98% 28% Hn 60% 10 0 0 0 0 0 0
98% 0 873 0 873 56480 320 24856 96716 0 0 1 98% 44% : 62% 0 0 0 0 0 0 0
97% 0 843 0 1021 56144 320 48404 101536 0 0 1 98% 42% H 57% 178 0 0 0 0 0 0
96% 0 842 0 851 56178 314 55542 26758 0 0 1 98% 15% Hf 58% 9 0 0 0 0 0 0
97% 0 888 0 888 57443 326 23033 79640 0 0 1 98% 45% : 69% 0 0 0 0 0 0 0
97% 0 844 0 844 54573 309 57296 103488 0 0 1 98% 43% H 65% 0 0 0 0 0 0 0
96% 0 852 0 858 57892 331 49632 67536 0 0 59s 98% 21% Hs 58% 6 0 0 0 0 0 0
97% 0 898 0 898 57384 322 19592 38108 0 0 59s 98% 64% : 60% 0 0 0 0 0 0 0
97% 0 852 0 858 57997 329 51556 101944 0 0 59s 98% 42% H 61% 6 0 0 0 0 0 0

ekashpureff · ‎2016-12-15

Yes, you're hammering your old 2240-4 with this test load.

CPU from systat is the greater of either the utilization of the busiest domain or the average utilization of all CPUs during the previous interval seconds.

You may want to look at sysstat -m to lsee load on all cores.

I noted that the performance is 57892 KB rather tah the Kb you referenced earlier.

Write workloads will be more CPU intensive than reads given the load on WAFL and RAID.

I hope this response has been helpful to you.

At your service,

Eugene E. Kashpureff, Sr.
Independent NetApp Consultant http://www.linkedin.com/in/eugenekashpureff
Senior NetApp Instructor, FastLane US http://www.fastlaneus.com/
(P.S. I appreciate 'kudos' on any helpful posts.)