[f-nsp] Unexpected CPU load on NetIron CER
Bardo Cornelissen
b.cornelissen at caveo.nl
Wed Oct 29 11:22:54 EDT 2014
Hi,
Thanks for the feedback. I have ip/icmp redirect disabled, that should not be the issue on its own. The only thing that may have some relation with it is that on the CER ip(v4) redirects cannot be disabled global (though it can and is by default for IPv6).
I actually use a virtual interface and also tried it directly on the physical interface but that makes no difference.
interface ve 11
port-name AMS-IX
ip address {ipv4-address}/21
no ip redirect
ipv6 address {ipv6-address}/64
ipv6 enable
ipv6 nd suppress-ra
!
interface ethernet 2/1
delay-link-event 20
enable
route-only
sflow forwarding
!
#sh ip int ve 11
….
ip icmp redirect: disabled
#sh ipv6 int ve 11
….
ICMP redirects are disabled
Kind regards,
Bardo Cornelissen
Caveo Internet BV
<http://www.caveo.nl/> http://www.caveo.nl
Van: Mat Perkins [mailto:mat at instavps.com]
Verzonden: woensdag 29 oktober 2014 15:55
Aan: Youssef Bengelloun-Zahr
CC: Bardo Cornelissen; foundry-nsp at puck.nether.net
Onderwerp: Re: [f-nsp] Unexpected CPU load on NetIron CER
I think the key here is:
no ip redirect
no ip icmp redirects
I have this running system wide, and I'm not on AMS-IX but I am on SIX
show cpu lp
SLOT #: LP CPU UTILIZATION in %:
in 1 second: in 5 seconds: in 60 seconds: in 300 seconds:
1: 1 1 1 1
3: 1 1 1 1
Mat
On Wed, Oct 29, 2014 at 3:09 AM, Youssef Bengelloun-Zahr <youssef at 720.fr> wrote:
Hello Bardo,
Running an MLXe-4 box (NI 5.4.0.d) connected to AMS-IX via 1G port, here is my config :
interface ethernet 3/3
port-name to PEERING : AMS-IX (EDGE-TC5-019 GE-2/1)
enable
load-interval 30
ip address blablablabla/21
no ip redirect
ip mtu 1500
ipv6 address blablablabla/64
ipv6 enable
ipv6 nd suppress-ra
sflow forwarding
No unusual high cpu-utilization :
09:08:34 GMT+00 Wed Oct 29 2014
SLOT #: LP CPU UTILIZATION in %:
in 1 second: in 5 seconds: in 60 seconds: in 300 seconds:
3: 4 5 5 6
Hope this helps.
Best regards.
2014-10-29 9:58 GMT+01:00 Bardo Cornelissen <b.cornelissen at caveo.nl>:
Hi,
Well that indeed solved the 50-60% load I experienced with "route-only"
configured on the interface. I'm still stuck 8% cpu-utilization on LP
though. Is anyone on this list connected to AMS-IX with a CER or MLX(e), not
experiencing this and willing to compare some interface configuration and
debug packet captures?
#sh cpu lp
09:37:37 GMT+01 Wed Oct 29 2014
SLOT #: LP CPU UTILIZATION in %:
in 1 second: in 5 seconds: in 60 seconds: in 300 seconds:
1: 8 8 8 8
2: 8 8 8 8
I'm seeing a lot of these when debugging:
**********************************************************************
[ppcr_rx_packet]: Packet received
Time stamp : 00 day(s) 09h 17m 48s:,
**********************************************************************
Packet size: 102, XPP reason code: 0x00000000
00: 0000 e40c 4010 e00b-1000 0000 0000 0000 FID = 0x0000
10: 3333 ff34 0001 0012-f29f 19c0 86dd 6c00 Offset = 0x10
20: 0000 0020 3aff 2001-07f8 0001 0000 0000 VLAN = 11(0x000b)
30: a502 5542 0001 ff02-0000 0000 0000 0000 CAM = 0x00000(L)
40: 0001 ff34 0001 8700-5787 0000 0000 2001 RX Qid = 6
50: 07f8 0001 0000 0000-a501 2634 0001 0101 CPU_CODE = 506
60: 0012 f29f 19c0 5555-5555 2634 0001 0101
70: 55ff ffff ffff ffff-5555 55e9 4003 0450
Pri CPU MON SRC PType US BRD DAV SAV DPV SV ER TXA SAS Tag MVID
7 1 0 2/1 c 0 1 0 0 0 0 0 0 0 0 0
2001:07f8:0001:0000:0000:a502:5542:0001 ->
ff02:0000:0000:0000:0000:0001:ff34:0001 ICMPv6
**********************************************************************
[ppcr_tx_packet] ACTION: Drop packet(reason: Route-only enabled(CV))
[ppcr_rx_packet]: Packet received
Time stamp : 00 day(s) 09h 17m 48s:,
**********************************************************************
Packet size: 102, XPP reason code: 0x00000000
00: 0000 e40c 4010 e00b-1000 0000 0000 0000 FID = 0x0000
10: 3333 ff96 0001 28c0-da85 bb00 86dd 6000 Offset = 0x10
20: 0000 0020 3aff fe80-0000 0000 0000 2ac0 VLAN = 11(0x000b)
30: da07 3285 bb00 ff02-0000 0000 0000 0000 CAM = 0x00000(L)
40: 0001 ff96 0001 8700-c35c 0000 0000 2001 RX Qid = 6
50: 07f8 0001 0000 0000-a504 3996 0001 0101 CPU_CODE = 506
60: 28c0 da85 bb00 5555-5555 3996 0001 0101
70: 55ff ffff ffff ffff-55c8 4003 0450 f9d2
Pri CPU MON SRC PType US BRD DAV SAV DPV SV ER TXA SAS Tag MVID
7 1 0 2/1 c 0 1 0 0 0 0 0 0 0 0 0
fe80:0000:0000:0000:2ac0:da07:3285:bb00 ->
ff02:0000:0000:0000:0000:0001:ff96:0001 ICMPv6
**********************************************************************
[ppcr_tx_packet] ACTION: Drop packet(reason: Route-only enabled(CV))
[ppcr_rx_packet]: Packet received
Time stamp : 00 day(s) 09h 17m 48s:,
**********************************************************************
Packet size: 102, XPP reason code: 0x00000000
00: 0000 e40c 4010 e00b-1000 0000 0000 0000 FID = 0x0000
10: 3333 ff50 0001 28c0-da85 bb00 86dd 6000 Offset = 0x10
20: 0000 0020 3aff fe80-0000 0000 0000 2ac0 VLAN = 11(0x000b)
30: da07 3285 bb00 ff02-0000 0000 0000 0000 CAM = 0x00000(L)
40: 0001 ff50 0001 8700-97e7 0000 0000 2001 RX Qid = 6
50: 07f8 0001 0000 0000-a505 6550 0001 0101 CPU_CODE = 506
60: 28c0 da85 bb00 5555-5555 6550 0001 0101
70: 55ff ffff ffff ffff-55da 4003 0450 f9d2
Pri CPU MON SRC PType US BRD DAV SAV DPV SV ER TXA SAS Tag MVID
7 1 0 2/1 c 0 1 0 0 0 0 0 0 0 0 0
fe80:0000:0000:0000:2ac0:da07:3285:bb00 ->
ff02:0000:0000:0000:0000:0001:ff50:0001 ICMPv6
**********************************************************************
Met vriendelijke groet,
Bardo Cornelissen
Caveo Internet BV
Ambachtsstraat 10
3176 PR Poortugaal
The Netherlands
http://www.caveo.nl
Tel: +31-10-5065520 <tel:%2B31-10-5065520>
Fax: +31-10-5065521 <tel:%2B31-10-5065521>
-----Oorspronkelijk bericht-----
Van: foundry-nsp [mailto:foundry-nsp-bounces at puck.nether.net] Namens Bardo
Cornelissen
Verzonden: dinsdag 28 oktober 2014 10:01
Aan: 'Lutz Knabenreich'; foundry-nsp at puck.nether.net
Onderwerp: Re: [f-nsp] Unexpected CPU load on NetIron CER
Hi,
Since I'm working on 5500c, that's worth a try ;-)
Thanks for your feedback!
Kind regards,
Bardo Cornelissen
Caveo Internet BV
http://www.caveo.nl
-----Oorspronkelijk bericht-----
Van: foundry-nsp [mailto:foundry-nsp-bounces at puck.nether.net] Namens Lutz
Knabenreich
Verzonden: dinsdag 28 oktober 2014 09:55
Aan: foundry-nsp at puck.nether.net
Onderwerp: Re: [f-nsp] Unexpected CPU load on NetIron CER
Hi,
> Has anyone seen similar behavior?
>
> IronWare : Version 5.5.0cT183 Copyright (c) 1996-2013 Brocade
Communications > Systems, Inc.
>
when we started to test the CER 2024F-4x in our environment we discovered
high cpu load as well, when using IPv6.
"sh cpu lp" was at unusual high levels:
09:15:15 GMT+01 Thu Sep 26 2013
SLOT #: LP CPU UTILIZATION in %:
in 1 second: in 5 seconds: in 60 seconds: in 300 seconds:
1: 67 67 67 67
2: 67 67 67 67
The root cause was found in a high amount of ipv6 discovery and
solliciatations.
In our specific case it was fixed in 5500d, and the issue was traced back to
the fix made for 459558 in 5500c.
Best regards
Lutz Knabenreich
--
HKN GmbH Tel: +49 211 875505-0
Lutz Knabenreich Fax: +49 211 87550555 <tel:%2B49%20211%2087550555>
Hochstadenstr. 5 http://www.hkn.de
47829 Krefeld mailto:knabenreich at hkn.de
Technik Standort:
In der Steele 29
40599 Duesseldorf
Eingetragen am Amtsgericht Krefeld HRB 12335
Geschaeftsfuehrer Lutz Knabenreich
_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
--
Youssef BENGELLOUN-ZAHR
_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/foundry-nsp/attachments/20141029/5ac8f501/attachment.html>
More information about the foundry-nsp
mailing list