Platform/versions:
# sh ver
kickstart: version 6.2(2)
system: version 6.2(2)
# sh mod
Mod Ports Module-Type Model Status
--- ----- ----------------------------------- ------------------ ----------
1 32 10 Gbps Ethernet Module N7K-M132XP-12 ok
2 48 1000 Mbps Optical Ethernet Module N7K-M148GS-11 ok
3 32 1/10 Gbps Ethernet Module N7K-F132XP-15 ok
4 32 1/10 Gbps Ethernet Module N7K-F132XP-15 ok
5 0 Supervisor Module-1X N7K-SUP1 ha-standby
6 0 Supervisor Module-1X N7K-SUP1 active *
I recently tried to add a couple of "ip dhcp relay address" statements to an SVI and received the following error message:
ERROR: Hardware programming failed. Reason: Tcam will be over used, please enable bank chaining and/or turn off atomic update
Studying this page I was able to determine, that I seemed to be hitting a 50% TCAM utilization limit on the F1 modules, which prevented atomic updates:
# show hardware access-list resource entries module 3
ACL Hardware Resource Utilization (Module 3)
--------------------------------------------
Instance 1, Ingress
--------------------
TCAM: 530 valid entries 494 free entries
IPv6 TCAM: 8 valid entries 248 free entries
Used Free Percent
Utilization
-----------------------------------------------------
TCAM 530 494 51.75
I was able to workaround it be disabling atomic updates:
hardware access-list update default-result permit
no hardware access-list update atomic
I understand, that with this config I am theoretically allowing ACL traffic during updates which shouldn't be allowed (alternately I could drop it), but that's not really my primary concern here.
First of all I need to understand why adding DHCP relays are apparently affecting my TCAM entry resources?
Second, I need to understand if there are other implications of disabling atomic updates, such as during ISSU?
Third, What are my options - if any - for planning the usage of the apparently relatively scarce resources on the F1 modules?
