CI: try to fix more networking flakes

There's a whole slew of networking-related flakes whose common
element seems to be improper use of curl. Fix those by:

  * add --retry --retry-connrefused; and/or
  * add -S ("show errors". Plain -s silences everything!); and/or
  * test exit status from curl; and/or
  * add wait_for_port after "podman run -d", to avoid races
  * log commands, to make debugging easier

Important note: wait_for_port() was not working with rootless
podman ports. Trivial proof:

  $ podman run -d --name foo -p 8192:80 \
      quay.io/libpod/testimage:20221018 \
      /bin/busybox-extras httpd -f -p 80
  $ grep :2000 /proc/net/tcp
  [no results]

Solution: use ss tool; it seems to handle this just fine.
There may be a better solution.

Oh, also, add -t1 to a podman restart, to shave 18s from test run.

Fixes: #20335 and, I think, a handful of others

Signed-off-by: Ed Santiago <[email protected]>

test/e2e/pod_infra_container_test.go

@@ -95,7 +95,7 @@ var _ = Describe("Podman pod create", func() {
 		session.WaitWithDefaultTimeout()
 		Expect(session).Should(ExitCleanly())
 
-		session = podmanTest.Podman([]string{"run", "--pod", podID, fedoraMinimal, "curl", "-s", "-f", "localhost:80"})
+		session = podmanTest.Podman([]string{"run", "--pod", podID, fedoraMinimal, "curl", "-s", "--retry", "2", "--retry-connrefused", "-f", "localhost:80"})
 		session.WaitWithDefaultTimeout()
 		Expect(session).Should(ExitCleanly())
 

test/system/500-networking.bats

@@ -59,9 +59,9 @@ load helpers.network
     is "$output" "$random_2" "exec cat index2.txt"
 
     # Verify http contents: curl from localhost
-    run curl -s $SERVER/index.txt
+    run curl -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt"
-    run curl -s $SERVER/index2.txt
+    run curl -s -S $SERVER/index2.txt
     is "$output" "$random_2" "curl 127.0.0.1:/index2.txt"
 
     # Verify http contents: wget from a second container
@@ -303,7 +303,7 @@ load helpers.network
     mac1="$output"
 
     # Verify http contents: curl from localhost
-    run curl -s $SERVER/index.txt
+    run curl -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt"
 
     # rootless cannot modify iptables
@@ -369,7 +369,7 @@ load helpers.network
     is "$output" "$mac2" "MAC address changed after podman network reload ($netname2)"
 
     # check that we can still curl
-    run curl -s $SERVER/index.txt
+    run curl -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt"
 
     # clean up the container
@@ -481,9 +481,10 @@ load helpers.network
             $IMAGE /bin/busybox-extras httpd -f -p 80
     cid=$output
 
-    # Verify http contents: curl from localhost
-    run curl --max-time 3 -s $SERVER/index.txt
-    is "$output" "$random_1" "curl 127.0.0.1:/index.txt"
+    # Verify http contents: curl from localhost. This is the first time
+    # connecting, so, allow retries until httpd starts.
+    run curl --retry 2 --retry-connrefused -s $SERVER/index.txt
+    is "$output" "$random_1" "curl $SERVER/index.txt"
 
     run_podman inspect $cid --format "{{(index .NetworkSettings.Networks \"$netname\").IPAddress}}"
     ip="$output"
@@ -505,16 +506,17 @@ load helpers.network
     run_podman exec $cid cat /etc/hosts
     assert "$output" !~ "$ip" "IP ($ip) should no longer be in /etc/hosts"
 
-    # check that we cannot curl (timeout after 3 sec)
-    run curl --max-time 3 -s $SERVER/index.txt
+    # check that we cannot curl (timeout after 3 sec). Fails with inconsistent
+    # curl exit codes, so, just check for nonzero.
+    run curl --max-time 3 -s -S $SERVER/index.txt
     assert $status -ne 0 \
            "curl did not fail, it should have timed out or failed with non zero exit code"
 
     run_podman network connect $netname $cid
     is "$output" "" "Output should be empty (no errors)"
 
     # curl should work again
-    run curl --max-time 3 -s $SERVER/index.txt
+    run curl --max-time 3 -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt should work again"
 
     # check that we have a new ip and mac
@@ -555,14 +557,14 @@ load helpers.network
     is "$output" "[${cid:0:12} $hostname]" "short container id and hostname in network2 aliases"
 
     # curl should work
-    run curl --max-time 3 -s $SERVER/index.txt
+    run curl --max-time 3 -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt should work"
 
     # disconnect the first network
     run_podman network disconnect $netname $cid
 
     # curl should still work
-    run curl --max-time 3 -s $SERVER/index.txt
+    run curl --max-time 3 -s -S $SERVER/index.txt
     is "$output" "$random_1" "curl 127.0.0.1:/index.txt should still work"
 
     # clean up
@@ -626,19 +628,26 @@ load helpers.network
 
         # Verify http contents again: curl from localhost
         # Use retry since it can take a moment until the new container is ready
-        run curl --retry 2 -s $SERVER/index.txt
-        is "$output" "$random_1" "curl 127.0.0.1:/index.txt after auto restart"
-
-        run_podman 0+w restart $cid
+        local curlcmd="curl --retry 2 --retry-connrefused -s $SERVER/index.txt"
+        echo "$_LOG_PROMPT $curlcmd"
+        run $curlcmd
+        echo "$output"
+        assert "$status" == 0 "curl exit status"
+        assert "$output" = "$random_1" "curl $SERVER/index.txt after auto restart"
+
+        run_podman 0+w restart -t1 $cid
         if ! is_remote; then
-            require_warning "StopSignal SIGTERM failed to stop container .* in 10 seconds, resorting to SIGKILL" \
+            require_warning "StopSignal SIGTERM failed to stop container .* in 1 seconds, resorting to SIGKILL" \
                             "podman restart issues warning"
         fi
 
         # Verify http contents again: curl from localhost
         # Use retry since it can take a moment until the new container is ready
-        run curl --retry 2 -s $SERVER/index.txt
-        is "$output" "$random_1" "curl 127.0.0.1:/index.txt after podman restart"
+        echo "$_LOG_PROMPT $curlcmd"
+        run $curlcmd
+        echo "$output"
+        assert "$status" == 0 "curl exit status"
+        assert "$output" = "$random_1" "curl $SERVER/index.txt after podman restart"
 
         run_podman rm -t 0 -f $cid
     done

test/system/700-play.bats

@@ -485,9 +485,10 @@ _EOF
     SERVER=http://127.0.0.1:$HOST_PORT
 
     run_podman run -d --name myyaml -p "$HOST_PORT:80" \
-    -v $PODMAN_TMPDIR/test.yaml:/var/www/testpod.yaml:Z \
-    -w /var/www \
-    $IMAGE /bin/busybox-extras httpd -f -p 80
+               -v $PODMAN_TMPDIR/test.yaml:/var/www/testpod.yaml:Z \
+               -w /var/www \
+               $IMAGE /bin/busybox-extras httpd -f -p 80
+    wait_for_port 127.0.0.1 $HOST_PORT
 
     run_podman kube play $SERVER/testpod.yaml
     run_podman inspect test_pod-test --format "{{.State.Running}}"

test/system/helpers.network.bash

@@ -241,6 +241,13 @@ function port_is_bound() {
         local proto="tcp"
     fi
 
+    # /proc/net/tcp is insufficient: it does not show some rootless ports.
+    # ss does, so check it first.
+    run ss -${proto:0:1}nlH sport = $port
+    if [[ -n "$output" ]]; then
+        return
+    fi
+
     port=$(printf %04X ${port})
     case "${address}" in
     *":"*)