snoop_cache_unresolved

  • 0
  • 1
  • Question
  • Updated 3 years ago
  • Answered
hi,
we have a stack with X460-G2 in summitstack.
EXOS is running in 15.6.3.1 v1563b1-patch1-8
The stacks has 6 members, with 2 uplink 10Gb/s to Cisco Core.
I think we have multicast problem : in logs, i have :
<Warn:Kern.Card.Warning> Slot-1: exosSnoop: snoop_cache_unresolved:851: snoop_cache_unresolved failed to send report
What does it mean ?
Anyone know that ? how to correct it ?

Thx for your help.
Regards,
Photo of herve colasuonno

herve colasuonno

  • 110 Points 100 badge 2x thumb

Posted 3 years ago

  • 0
  • 1
Photo of Bill Stritzinger

Bill Stritzinger, Alum

  • 6,036 Points 5k badge 2x thumb
What is the Multicast configuration in this network?  Is there PIM-SM?  If so can you provide the PIM configuration from the XOS side?
Photo of herve colasuonno

herve colasuonno

  • 110 Points 100 badge 2x thumb
On cisco core, there pim sparse-dense mode on all vlan on each vrf. The RP for each vrf, is the loopback of the vrf.
On exos, nothing for multicast, just :
#
# Module mcmgr configuration.
#
enable igmp snooping forward-mcrouter-only vr VR-Default
configure igmp snooping filters per-vlan
configure forwarding ipmc lookup-key group-vlan
configure mcast ipv4 cache timeout 90

The configure igmp snooping filters per-vlan is new since this afternoon. I know it's per port by default.

Thx for you help.
Photo of Bill Stritzinger

Bill Stritzinger, Alum

  • 6,036 Points 5k badge 2x thumb
Herve, 

What are the other switches in the stack and what version of firmware are you running.. can you post the output of "show ver" and "show slot" ?  Looks like you are exceeding queue space on the mcast and the message indicates the software is trying to protect the integrity of the CPU in the master of the stack. Are the mcast queues very large, I assume you are running a very large multicast setup?  You can see the queue "show mcast cache" - send the last page of that too..  Do you by chance have a TAC case open for this?

Bill
Photo of Bill Stritzinger

Bill Stritzinger, Alum

  • 6,036 Points 5k badge 2x thumb
Also .. "show igmp snooping detail" - Are there any loops possibly on the network? This message indicated there is some device that is consuming the cache, it is only thrown if there are 2048 packets to be processed by the snooping module.  Seems to be a device not playing nice, are there any EAPS rings or anything else in the environment?
Photo of herve colasuonno

herve colasuonno

  • 110 Points 100 badge 2x thumb
Hi Bill,
here comes :
Slot-1 Extreme-Stack-3.1 # sh ver
Slot-1      : 800540-00-04 1445N-41633 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-2      : 800540-00-04 1445N-41632 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-3      : 800540-00-04 1344N-40505 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-4      : 800540-00-04 1344N-40504 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-5      : 800540-00-04 1344N-40455 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-6      : 800540-00-04 1445N-41636 Rev 4.0 BootROM: 2.0.2.3    IMG: 15.6.3.1
Slot-7      :
Slot-8      :
XGM3S-2sf-A-1: 800450-00-04 1434N-42176 Rev 4.0
XGM3S-2sf-A-6: 800450-00-04 1434N-42181 Rev 4.0
SS-V80-B-1  : 800385-00-03 1439N-42095 Rev 3.0
SS-V80-B-2  : 800385-00-03 1439N-42094 Rev 3.0
SS-V80-B-3  : 800385-00-03 1439N-42096 Rev 3.0
SS-V80-B-4  : 800385-00-03 1439N-42099 Rev 3.0
SS-V80-B-5  : 800385-00-03 1439N-42098 Rev 3.0
SS-V80-B-6  : 800385-00-03 1439N-42076 Rev 3.0

Image   : ExtremeXOS version 15.6.3.1 v1563b1-patch1-8 by release-manager
          on Thu Oct 1 10:42:28 EDT 2015
BootROM : 2.0.2.3
Diagnostics : 6.3

Slot-1 Extreme-Stack-3.2 # sh slot
Slots    Type                 Configured           State       Ports
--------------------------------------------------------------------
Slot-1   X460-48x             X460-48x             Operational   54
Slot-2   X460-48x             X460-48x             Operational   54
Slot-3   X460-48x             X460-48x             Operational   54
Slot-4   X460-48x             X460-48x             Operational   54
Slot-5   X460-48x             X460-48x             Operational   54
Slot-6   X460-48x             X460-48x             Operational   54
Slot-7                                             Empty          0
Slot-8                                             Empty          0

sh mcast cache show many pages, because we have so many vlans in networks (about 100)

sh igmp snooping detail shows many pages. The last one is :
VLAN E3-VLAN251-PREDIS-GICS (251) Snooping=Enabled
    Group           Sender          Age
    224.0.0.13      195.220.24.126  13
    1 Incoming multicast streams

    Port   Host            Subscribed        Age    Group-Limit
    1:49   195.220.24.126  All Groups        13     0


VLAN E3-VLAN252-ENSEIG-NOMADES (252) Snooping=Enabled
    Group           Sender          Age
    224.0.0.13      195.220.24.190  76
    1 Incoming multicast streams

    Port   Host            Subscribed        Age    Group-Limit
    1:49   195.220.24.190  All Groups        34     0


VLAN CROUS-VLAN633    (633) Snooping=Enabled


VLAN PHITEM-ADMIN     (350) Snooping=Enabled
    Group           Sender          Age
    224.0.0.13      193.48.254.30   8
    1 Incoming multicast streams

    Port   Host            Subscribed        Age    Group-Limit
    4:35   193.48.254.27   239.255.255.250   33     0
    1:49   193.48.254.30   All Groups        8      0


VLAN PHITEM-ETU       (351) Snooping=Enabled
    Port   Host            Subscribed        Age    Group-Limit
    1:49   193.48.255.254  All Groups        40     0


    s = static igmp member
So if i'm on the limist of hars multicast, how can i block multicast in non wanted vlans/port ?

Thx for you hlep
Photo of Bill Stritzinger

Bill Stritzinger, Alum

  • 6,036 Points 5k badge 2x thumb
I would assume you have a ton of vlans..?  To disable snooping on a per vlan basis use the command "disable igmp snooping vlan <name>" -  I would look for the vlan that has the most streams or ones that you do not need.
Photo of herve colasuonno

herve colasuonno

  • 110 Points 100 badge 2x thumb
That's what i wanted to do in the first time...
But what are the consequences ? If i disable igmp on many (or all) vlans, what happend when a computer will ask multicast with Bonjour, or Up'n'p ? Does it flood all the stack to reach a multicast router on the cisco core ? And at last, is igmp necessary on L2 network ?
So sorry for thoses questions, but i'm not familliar with multicast.

Thx ,
Photo of EtherMAN

EtherMAN, Embassador

  • 7,370 Points 5k badge 2x thumb
If you disable IGMP snooping all mcast will be treated like a broadcast packet and send to all ports in each vlan.  We run 1800 or so vlans in our Metro network and only have igmp snooping enable at edge switches as a rule.  Only exception to that is we do have one vlan in our core that is IPTV traffic and at peak it hits 1.4 Gbs so we do keep it enabled so we don't overrun any one gig aggregation links coming off the core.  By have it enabled it only runs about 650 Mbs in the core.  So to answer your question is two fold.  How much mcast traffic is in each vlan?  If not that much then in our experience with a lot of vlans in one switch it is better to disable IGMP snooping globally and go back and enable it as needed.  I would guess since you are running in Spars Mode there is not much Mcast traffic in any one single vlan ??  
Photo of herve colasuonno

herve colasuonno

  • 110 Points 100 badge 2x thumb
Hi Etherman,

i don't know what is the muticast traffic. We don't do multicast deployement actually. So, only windows computers, or VDI products, does multicast. But i don't know how they do multicast. I think with U-p'n'p, bonjour, or something like that.

The only things, is when i hace the log message snoo_cache_unresolved, one of the slot in my stack takes 90 to 100% CPU. Then the slot reboot or failed, and the stack becomes unstable.

So i think i have multicast problems on my networks, but i don't know how to treat it. Only because leak of CPU on SummitStack X460-G2. In our last campus, we have Cisco 2950 or 2960, with same architecture, without any multicast problem.

It seems like the satck wants to treat multicast traffic, but i use the stack in L2 only. Ans maybe the stack wants to do some L3 with multicast.

Or i'm totally wrong. Actually, the most likely is multicast problem, maybe multicast loop. But i'm not sure. Only leak of cpu on stacks... And i don't how how can i have multicast loop, without any multicast configuration !
(Edited)