[j-nsp] MX80 no more hash-key option in 12.2?

[Paul Vlaar]

On 23/10/12 6:37 PM, Olivier Benghozi wrote:
> I had previously (like you) the hash-key stuff configured on some
> MX80 gear, and removed it after reading David Roy's answer.
> 
> I can confirm that the removal of this stuff left the tfeb in a
> strange state (running 11.4R5), with some funny logs at the commit
> (tfeb0 jnh_loadbalance_hashkey(538): Unsupported Hash Mode:4,
> Flags:0x1 received, tfeb0 jnh_loadbalance_hashkey(538): Unsupported
> Hash Mode:11, Flags:0x0 received), and a buncha "No" instead of the
> "Yes" I was looking for in the "show jnh lb" results.
> 
> I added some enhanced-hash-key options, commited, removed them,
> commited, and the Hash status was finally fine with the proper
> default Yes at the proper places... A bit rough and unpolished I
> guess :)

Hah, interesting. I get similar weirdness:

no family inet6 config under [forwarding-options enhanced-hash-key] :

GOT:            IIF-V6: No
GOT:          SPORT-V6: No
GOT:          DPORT-V6: No
GOT:     TRAFFIC_CLASS: No

[edit forwarding-options enhanced-hash-key]
mx80# set family inet6 no-source-port

GOT:            IIF-V6: No
GOT:          SPORT-V6: No
GOT:          DPORT-V6: Yes
GOT:     TRAFFIC_CLASS: No

So not setting the source port to "No" causes the destination port
setting to flip to Yes!

[edit forwarding-options enhanced-hash-key]
mx80# set family inet6 no-destination-port

GOT:            IIF-V6: No
GOT:          SPORT-V6: No
GOT:          DPORT-V6: No
GOT:     TRAFFIC_CLASS: No

[edit forwarding-options enhanced-hash-key]
mx80# deactivate family inet6 no-destination-port

[edit forwarding-options enhanced-hash-key]
mx80# deactivate family inet6 no-source-port

GOT:            IIF-V6: No
GOT:          SPORT-V6: Yes
GOT:          DPORT-V6: Yes
GOT:     TRAFFIC_CLASS: No

[edit forwarding-options enhanced-hash-key]
mx80# delete
Delete everything under this level? [yes,no] (no) yes

GOT:            IIF-V6: No
GOT:          SPORT-V6: Yes
GOT:          DPORT-V6: Yes
GOT:     TRAFFIC_CLASS: No

This definitely seems like a bug. Worth submitting to Juniper TAC?

	~paul