On 06/27/2012 02:28 PM, Dan McDonald wrote:
>
> On Jun 27, 2012, at 8:23 AM, Sašo Kiselkov wrote:
>
>> On 06/27/2012 01:47 PM, Udo Grabowski (IMK) wrote:
>>> ip_squeue_fanout=1 ?
>>
>> Thanks for the tip. Just tried it even with a reboot and it didn't help.
>> My cpu23 is still being hammered hard. Back in S10 there used to be a
>> ip_soft_rings_cnt variable, but that's gone in Illumos.
>
> Are you running this on a VLAN by any chance? If you are, it's possible
> you're being bitten by Illumos #918.
There are some similarities to #918, but here's the strange thing:
# mdb -k
> ::mac_srs -rcv
CPU_COUNT FANOUT_CPU_COUNT
ADDR LINK_NAME (CPU_LIST) (CPU_LIST)
ffffff2203962680 igb3 2 1
(17,18) (17)
ffffff2203963340 igb3 2 1
(15,16) (15)
ffffff21ec06e980 igb1 2 1
(0f,10) (0f)
ffffff21ec06f640 igb2 2 1
(11,12) (11)
ffffff21ec070300 igb2 2 1
(13,14) (13)
ffffff21ec074340 igb1 2 1
(0d,0e) (0d)
The CPU that's being floored my case is cpu23 (0x17 hexa above on igb3 -
the one receiving the data). Why did it elect to only do fanout on a
single CPU? I can also see cpu24 (0x18) being somewhat thrashed, but far
less than cpu23 (about half). The rest of the CPUs is pretty much idle
(90-95% idle).
Cheers,
--
Saso
-------------------------------------------
illumos-discuss
Archives: https://www.listbox.com/member/archive/182180/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4
Powered by Listbox: http://www.listbox.com
