Background:
An evaluation system setup, with two hubs (H1 (Balance 380 HW6), H2 (Balance 380 HW5)) and a small number of leaves (L1 (HD2), L2 (HD2), L3 (Transit), L4 (Balance 30), L5 (Balance One Core)).
Initial configuration:
Set up using InControl2.
H1 and H2 are connected.
L1, L2 and L3 are all connected to both H1 and H2 stars.
L4 is connected to H1.
Everything is copacetic and functioning well.
Disruption:
Adding L5 as a leaf for both H1 and H2.
Result:
- L5 connects to both H1 and H2 stars.
- The H2-L3 connection drops and IC2 reports it to be hung in the “start” state. The H1-L3 connection is unaffected.
Rebooting H2 and L3 did not resolve the issue.
Question 1:
Why would adding L5 kill the L3 connection?
Then:
- We removed L3 from the H2-centered star, and created a point-to-point of L3-L4.
The connection came up nicely. - We removed that connection and added an H2-L3 point-to-point connection (recreating the H2-centered star in an inelegant way
).
The connection came up nicely - We then removed that connection, and added L3 back to the H2-centered star.
The connection came up nicely
The last step above essentially recreated the same configuration that failed earlier.
Question 2:
What might have caused the problem in the first place, and what could have made it go away after the little dance afterwards?
We are a bit troubled by the experienced lack of predictability (or understanding on our part) of the effect of adding a node to a network.