frequent failure in 150_connect_forget_test

[rade]

e.g. here

---= Running 150_connect_forget_test.sh =---
Connecting and forgetting routers after launch
7e52b95f812d0bf678250a792b650fac3be0950e5fcd7a2a5a1f8ab9479129e7
fedc53aa1431eb10bcd877f3d19b5aefc7ad1cd3f7da28c6e81755ac99543d0a
fcb9ef937002787a01ba8a53c3e23d26c9e339c398e3b7292e662f738be577de
1c0712d8f0f27a8e659edef20370e0da8eb8db67946aafd08b60effb060ef201
test #6 "exec_on host1-1323-1.us-central1-a.positive-cocoa-90213 c1 ping -nq -W 1 -c 1 10.2.1.7" failed:
    program terminated with code 1 instead of 0
1 of 11 tests failed in 17.464s.

I suspect it's just timing; we've just connected to another peer and odds are the topology and routing info hasn't fully updated yet.

[awh]

I've seen this a couple of times locally with 110_encryption_test.sh, not quite enough times to make an issue of it yet but I'll keep an eye on it.

[rade]

I've seen this a couple of times locally with 110_encryption_test.sh

So have I, but that's a separate issue. Though possibly the same cause.

[awh]

@bboreham do you have any thoughts on whether #1052 fixes this too?

[bboreham]

See #1052 (comment)

[rade]

I haven't seen this fail for a long time. Let's assume #1052 fixed it.

[bboreham]

Failed on my local PC today.

[bboreham]

I see that the two peers can disconnect because they have inconsistent IPAM ring data.
Suggest starting with --init-peer-count=2.

[tomwilkie]

The test does the following:

1. start 2 routers, not connected
2. start 2 containers (1 on each), with manually assigned IPs
3. connect the 2 routers

The change which stops this working (#1200) is that we now check manually assigned IPs with IPAM, which means a ring is being created at 2 and 3 is failing.

[bboreham]

Is it wrong to add the --init-peer-count=2? I guess it's unnecessary given your change, but I'm thinking we should follow our own rules.