Reset Search
 

 

Article

X440 CPU high due to "bcmRx" process around 45-50%

« Go Back

Information

 
TitleX440 CPU high due to "bcmRx" process around 45-50%
Symptoms
  • The Summit x440 causing high CPU due to "bcmRx" process.
  • The "top" output shows "bcmRx" process around 45-50%
  • Check the output of "debug hal show congestion" to verify the CPU congestion is present or not. 
  • Collect the tcp dump capture and verify the large amount of packets hitting the CPU.

->debug hal show forwarding distributions system
L3 Hash Table Utilization:
  Used/Available: 512/512


L3 Hash Table Bucket Distribution:
  Total: 64 half-buckets
    [ 0 entries]:     0 half-buckets
    [ 1   entry]:     0 half-buckets
    [ 2 entries]:     0 half-buckets
    [ 3 entries]:     0 half-buckets
    [ 4 entries]:     0 half-buckets
    [ 5 entries]:     0 half-buckets
    [ 6 entries]:     0 half-buckets
    [ 7 entries]:     0 half-buckets
    [ 8 entries]:    64 half-buckets

  • Check the LPM table to confirm if it is not full or not.
  • Please see the below lab output: 

>show iproute reserved-entries statistics
                      |-----In HW Route Table-----|   |-------In HW L3 Hash Table------|
                       # Used Routes   # IPv4 Hosts   IPv4   IPv4  IPv6    IPv4    IPv6
Slot  Type              IPv4   IPv6    Local Remote   Local  Rem.  Local   MCast   MCast
----  --------------- ------- ------  ------ ------   -----  ----- -----  ------  ------
1     X440-24x-10G         13      0      48      0     314      0     0      99       0

Theoretical maximum for each resource type:
X440                       32     16      64     64     509    512   256 *   256 *   128

  • Check the output of "show forwarding configuration" to verify the proper hashing algorithm "crc32" with "dual-hash" is on.
  • Also, verify the configured Lookup-key.

->show forwarding configuration
L2 and L3 Forwarding table hash algorithm:
    Configured hash algorithm:              crc32
    Current hash algorithm:                 crc32
L3 Dual-Hash configuration:
    Configured setting:                     on
    Current setting:                        on
    Dual-Hash Recursion Level:              1
Hash criteria for IP unicast traffic for L2 load sharing and ECMP route sharing
    Sharing criteria:                       L3_L4
IP multicast:
    Group Table Compression:                on
    Local Network Forwarding:               slow-path
    Lookup-Key:                             (SourceIP, GroupIP, VlanId)
Switch Settings:
    Switching mode:                         store-and-forward
L2 Protocol:
    Fast convergence:                       on
Fabric Flow Control:
    Fabric Flow Control:                    auto

 

Environment
  • EXOS all
  • Summit X440
Cause
  • It could be due to large amount of "known" unicast traffic hitting the CPU and a lot of TCP retransmissions between devices in the specific VLAN's.
Resolution
  • Try to change the lookup-key to "group-vlan" for better usage of the L3 hash table using the below command - 
    "Configure forwarding ipmc lookup-key group-vlan"
  • Post this, verify the distribution of hashed entries in the "l3 hash table" and the CPU utilization for "bcmRx" process.
  • If issues still persist, please seek assistance from GTAC.
Additional notes

Feedback

 

Was this article helpful?


   

Feedback

Please tell us how we can make this article more useful.

Characters Remaining: 255