Repository navigation
Conversation
troglobit
left a comment
There was a problem hiding this comment.
Minor comments, mostly for internal discussion.
| deviation "/rt:routing/rt:control-plane-protocols/rt:control-plane-protocol/ospf:ospf/ospf:auto-cost/ospf:reference-bandwidth" { | ||
| deviate add { | ||
| default "100000"; | ||
| } |
There was a problem hiding this comment.
This needs the eyes of an expert, @jovatn how will this affect existing Styx installations?
There was a problem hiding this comment.
Ah, I see now that the default is the same as the previous hard-coded value in Frr. So less critical, but maybe we should change the default? What do you think @jovatn ...
There was a problem hiding this comment.
I'm a bit in a hurry now, but lets discuss next week if you think we should change.
We have not discussed this, but my main intention was to use FRR defaults as Infix defaults (see initial description)
- Then we should not need migration scripts
- And I believe that 100Gbit/s is a better reference bandwidth for the future than 100Mbit/s (or whatever the standard may say).
Still, it is wise to lookup existing styx installations, although I believe/hope that adjustments (if needed) be done on appropriate repo than on mainline Infix.
5edf54b to
22355ea
Compare
All routers in an OSPF domain must use the same reference bandwidth to agree on link costs, and other vendors default to other values than Frr. Until now the value was fixed at the Frr default. Enable the ietf-ospf auto-cost feature for OSPFv2 and OSPFv3. The reference-bandwidth leaf defaults to 100000 Mbit/s, the value used so far, so link costs do not change on upgrade and no migration is needed. Frr can not turn auto-cost off, so the enabled leaf defaults to true and false is rejected. Resolves: #1576 Signed-off-by: Ejub Sabic <ejub1946@outlook.com>
To check that the reference bandwidth matches across the domain it has to be readable. Frr has no JSON for it, so read it from the daemon's running-config, which lists it only when it differs from the default. Report it for OSPFv2 and OSPFv3, and show it in 'show ip ospf' and 'show ipv6 ospf'. Signed-off-by: Ejub Sabic <ejub1946@outlook.com>
Verify that the interface cost follows the reference bandwidth, for OSPFv2 and OSPFv3. The expected cost is derived from the link speed, so the test works on hardware and on virtual links of unknown speed. On a 1 Gbit/s link the reference bandwidths 100000, 10000 and 1000 Mbit/s give the costs 100, 10 and 1. Signed-off-by: Ejub Sabic <ejub1946@outlook.com>
Explain why the reference bandwidth must match across the domain, how the interface cost is calculated, and where to read the value. Signed-off-by: Ejub Sabic <ejub1946@outlook.com>
22355ea to
a46aefa
Compare
Description
All OSPF routers in a domain need the same auto-cost reference bandwidth
to agree on link costs. Other vendors use other defaults than Frr, so
the value has to be configurable. Resolves #1576.
The
auto-costcontainer from ietf-ospf is now supported, for OSPFv2and OSPFv3:
reference-bandwidthdefaults to 100000 Mbit/s, the Frr default usedso far. Link costs do not change on upgrade, so no migration script
is needed.
enabledis always true. Frr has no way to turn auto-cost off, sosetting it to false is rejected. An explicit interface
coststilloverrides the calculated one.
shown by
show ip ospfandshow ipv6 ospf.The new test
ospf_auto_costruns for both OSPF versions. It sets thereference bandwidth on both routers and checks the interface cost. On
a 1 Gbit/s link, 100000, 10000 and 1000 Mbit/s give the costs 100, 10
and 1.
Checklist
Tick relevant boxes, this PR is-a or has-a: