tests: Fix "interconnection" ic test.

This test started recently [1] to become flaky (since remote ports are
related ports).
Note that the test should in fact always fail (and not only randomly),
but due to some (independent) issues in CHECK_RELATED_PORTS_AFTER_RECOMPUTE
this is not always visible.
The CHECK_RELATED_PORTS_AFTER_RECOMPUTE macro issue will be
fixed separately.

[1] ba2db3b ("northd: Fix broadcast of all traffic within a transit spine switch.")

Fixes: ba2db3b ("northd: Fix broadcast of all traffic within a transit spine switch.")
Signed-off-by: Xavier Simonart <[email protected]>
Signed-off-by: Numan Siddique <[email protected]>

Author: simonartxavier

tests/ovn-ic.at

@@ -1721,8 +1721,24 @@ for s_az in $(seq 1 $n_az); do
 done
 OVS_WAIT_UNTIL([check_packets], [$at_diff -F'^---' expected received])
 
+# ts[i] become local datapaths in hv[az] right before we lrp-set the
+# gateway-chassis on lrp-lr$az-$i-ts$i.
+# For ts1 & ts2, there are recomputes after the gateway-chassis has been set
+# (e.g. when some tunnel_keys get added), and hence they get removed from
+# local datapaths.
+# However this is not the case for ts3: there is no recompute, and local
+# datapaths are usually not removed in I+P. Ignore this.
+# Also, ts3 remaining a local datapath, and lsp-ts3-lr[x]-3 being remote port, it
+# is a related port. Ignore this also
+ovn_as az1 OVN_CLEANUP_SBOX([hv1], [], ["lsp-ts3-lr2-3
+lsp-ts3-lr3-3"], [ts3])
+ovn_as az2 OVN_CLEANUP_SBOX([hv2], [], ["lsp-ts3-lr1-3
+lsp-ts3-lr3-3"], [ts3])
+ovn_as az3 OVN_CLEANUP_SBOX([hv3], [], ["lsp-ts3-lr1-3
+lsp-ts3-lr2-3"], [ts3])
+
 for az in `seq 1 $n_az`; do
-    OVN_CLEANUP_SBOX([hv$az])
+    ovn_as az$az
     OVN_CLEANUP_SBOX([gw$az])
     OVN_CLEANUP_AZ([az$az])
 done