Batfish: aggregate with no bound interfaces has bandwidth 0

[dhalperi]

Not null, which breaks downstream pipeline.

Fix #4124

[batfish-bot]

This change is 

[dhalperi]

Reviewable status: 0 of 3 files reviewed, 1 unresolved discussion (waiting on @progwriter)

a discussion (no related file):
Some thoughts:

• PC1 itself does not appear in the VI Model. That's bad, right?
• Is this the right fix?

---

[codecov]

Codecov Report

Merging #4125 into master will decrease coverage by 0.1%.
The diff coverage is 88.88%.

@@             Coverage Diff              @@
##             master    #4125      +/-   ##
============================================
- Coverage     75.23%   75.13%   -0.11%     
+ Complexity    24877    24656     -221     
============================================
  Files          2030     2026       -4     
  Lines         99493    98628     -865     
  Branches      11796    11723      -73     
============================================
- Hits          74857    74106     -751     
+ Misses        19293    19194      -99     
+ Partials       5343     5328      -15

Impacted Files	Coverage Δ	Complexity Δ
...atfish/src/main/java/org/batfish/main/Batfish.java	75.4% <88.88%> (+0.02%)	298 <2> (ø)	⬇️
.../java/org/batfish/symbolic/state/ExitsNetwork.java	50% <0%> (-50%)	1% <0%> (-1%)
...fish/symbolic/state/NodeInterfaceExitsNetwork.java	66.66% <0%> (-33.34%)	1% <0%> (-1%)
...eachability/transition/RemoveSourceConstraint.java	53.33% <0%> (-13.34%)	5% <0%> (-1%)
.../src/main/java/org/batfish/datamodel/flow/Hop.java	53.33% <0%> (-6.67%)	5% <0%> (-1%)
...c/main/java/org/batfish/dataplane/rib/RibTree.java	89.79% <0%> (-4.09%)	27% <0%> (-1%)
...rc/main/java/org/batfish/datamodel/flow/Trace.java	87.09% <0%> (-3.23%)	11% <0%> (-2%)
...rc/main/java/org/batfish/bddreachability/Edge.java	67.64% <0%> (-2.95%)	11% <0%> (-1%)
...sh/bddreachability/SessionCreationNodeVisitor.java	39.39% <0%> (-2.64%)	13% <0%> (-16%)
...h/bddreachability/ReversePassOriginationState.java	41.17% <0%> (-2.49%)	14% <0%> (-17%)
... and 15 more

[corinaminer]

Reviewable status: 0 of 3 files reviewed, 1 unresolved discussion (waiting on @dhalperi and @progwriter)

a discussion (no related file):

Previously, dhalperi (Dan Halperin) wrote…

Some thoughts:

• PC1 itself does not appear in the VI Model. That's bad, right?
• Is this the right fix?

If i understand portchannel interfaces and subinterfaces correctly, i believe you're right that PC1 ought to appear. However, CiscoConfiguration relies on it being explicitly defined; see also this test config.

I think we want to automatically create a VI portchannel interface when a subinterface is defined but its parent isn't (i.e., at the point in CiscoConfiguration linked above). Then i don't think it would be possible for an AGGREGATE_CHILD interface to have no BIND dependency (at least for Cisco portchannel interfaces).

---

[progwriter]

Reviewed 3 of 3 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @dhalperi)

[progwriter]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @dhalperi)

a discussion (no related file):
Another issue is why is ospf cost being computed in the first place. This interface should be down.
Based on offline discussion, it appears that conversion should create dependencies on non-existent interfaces and code in snapshot post-processing should turn down interfaces with bind dependencies that point to non-existent interfaces. Some other code (such as l1 logical topology computation) might need to be special cased as well.

---