0ct0pu5/moby
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

De-flake TestSwarmClusterRotateUnlockKey... again... maybe?

Browse source 
This hopefully makes the test less flakey (or removes any flake that
would be caused by the test itself).

1. Adds tail of cluster daemon logs when there is a test failure so we
   can more easily see what may be happening
2. Scans the daemon logs to check if the key is rotated before
   restarting the daemon. This is a little hacky but a little better
   than assuming it is done after a hard-coded 3 seconds.
3. Cleans up the `node ls` check such that it uses a poll function

Signed-off-by: Brian Goff <cpuguy83@gmail.com>
(cherry picked from commit fbdc02534a)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This commit is contained in:
Brian Goff 2024-01-02 20:29:32 +00:00 • committed by Sebastiaan van Stijn
parent 811585331c
commit 97433635a6
No known key found for this signature in database
GPG key ID: 76698F39D527CE8C
3 changed files with 71 additions and 54 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
integration-cli/check_test.go
View file

@ -576,6 +576,9 @@ func (s *DockerSwarmSuite) TearDownTest(c *testing.T) {
s.daemonsLock.Lock()
for _, d := range s.daemons {
if d != nil {
if c.Failed() {
d.TailLogsT(c, 100)
}
d.Stop(c)
d.Cleanup(c)
}

111
integration-cli/docker_cli_swarm_test.go
View file

@ -27,6 +27,7 @@ import (
"github.com/docker/docker/libnetwork/driverapi"
"github.com/docker/docker/libnetwork/ipamapi"
remoteipam "github.com/docker/docker/libnetwork/ipams/remote/api"
testdaemon "github.com/docker/docker/testutil/daemon"
"github.com/moby/swarmkit/v2/ca/keyutils"
"github.com/vishvananda/netlink"
"gotest.tools/v3/assert"
@ -1223,7 +1224,10 @@ func (s *DockerSwarmSuite) TestSwarmJoinPromoteLocked(c *testing.T) {
poll.WaitOn(c, pollCheck(c, d3.CheckLocalNodeState, checker.Equals(swarm.LocalNodeStateActive)), poll.WithTimeout(time.Second))
}
const swarmIsEncryptedMsg = "Swarm is encrypted and needs to be unlocked"
func (s *DockerSwarmSuite) TestSwarmRotateUnlockKey(c *testing.T) {
ctx := context.TODO()
d := s.AddDaemon(c, true, true)
outs, err := d.Cmd("swarm", "update", "--autolock")
@ -1242,12 +1246,16 @@ func (s *DockerSwarmSuite) TestSwarmRotateUnlockKey(c *testing.T) {
d.RestartNode(c)
assert.Equal(c, getNodeStatus(c, d), swarm.LocalNodeStateLocked)
outs, _ = d.Cmd("node", "ls")
assert.Assert(c, strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
cmd := d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(unlockKey)
result := icmd.RunCmd(cmd)
unlock := func(d *daemon.Daemon, key string) *icmd.Result {
cmd := d.Command("swarm", "unlock")
cmd.Stdin = strings.NewReader(key)
return icmd.RunCmd(cmd)
}
outs, _ = d.Cmd("node", "ls")
assert.Assert(c, strings.Contains(outs, swarmIsEncryptedMsg), outs)
result := unlock(d, unlockKey)
if result.Error == nil {
// On occasion, the daemon may not have finished
// rotating the KEK before restarting. The test is
@ -1257,13 +1265,16 @@ func (s *DockerSwarmSuite) TestSwarmRotateUnlockKey(c *testing.T) {
// restart again, the new key should be required this
// time.
time.Sleep(3 * time.Second)
// Wait for the rotation to happen
// Since there are multiple rotations, we need to wait until for the number of rotations we are currently on to be reflected in the logs
// This is a little janky... its depending on specific log messages AND these are debug logs... but it is the best we can do for now.
matcher := testdaemon.ScanLogsMatchCount(testdaemon.ScanLogsMatchString("successfully rotated KEK"), i+1)
poll.WaitOn(c, d.PollCheckLogs(ctx, matcher), poll.WithDelay(3*time.Second), poll.WithTimeout(time.Minute))
d.Restart(c)
d.RestartNode(c)
cmd = d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(unlockKey)
result = icmd.RunCmd(cmd)
result = unlock(d, unlockKey)
}
result.Assert(c, icmd.Expected{
ExitCode: 1,
@ -1271,28 +1282,20 @@ func (s *DockerSwarmSuite) TestSwarmRotateUnlockKey(c *testing.T) {
})
outs, _ = d.Cmd("node", "ls")
assert.Assert(c, strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
cmd = d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(newUnlockKey)
icmd.RunCmd(cmd).Assert(c, icmd.Success)
assert.Assert(c, strings.Contains(outs, swarmIsEncryptedMsg), outs)
unlock(d, newUnlockKey).Assert(c, icmd.Success)
assert.Equal(c, getNodeStatus(c, d), swarm.LocalNodeStateActive)
retry := 0
for {
checkNodeLs := func(t poll.LogT) poll.Result {
// an issue sometimes prevents leader to be available right away
outs, err = d.Cmd("node", "ls")
if err != nil && retry < 5 {
if strings.Contains(outs, "swarm does not have a leader") {
retry++
time.Sleep(3 * time.Second)
continue
}
out, err := d.Cmd("node", "ls")
if err != nil {
return poll.Continue("error running node ls: %v: %s", err, out)
}
assert.NilError(c, err)
assert.Assert(c, !strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
break
return poll.Success()
}
poll.WaitOn(c, checkNodeLs, poll.WithDelay(3*time.Second), poll.WithTimeout(time.Minute))
unlockKey = newUnlockKey
}
@ -1308,6 +1311,7 @@ func (s *DockerSwarmSuite) TestSwarmClusterRotateUnlockKey(c *testing.T) {
if runtime.GOARCH == "ppc64le" {
c.Skip("Disabled on ppc64le")
}
ctx := context.TODO()
d1 := s.AddDaemon(c, true, true) // leader - don't restart this one, we don't want leader election delays
d2 := s.AddDaemon(c, true, true)
@ -1329,15 +1333,23 @@ func (s *DockerSwarmSuite) TestSwarmClusterRotateUnlockKey(c *testing.T) {
d2.RestartNode(c)
d3.RestartNode(c)
unlock := func(d *daemon.Daemon, key string) *icmd.Result {
cmd := d.Command("swarm", "unlock")
cmd.Stdin = strings.NewReader(key)
return icmd.RunCmd(cmd)
}
const swarmIsEncryptedMsg = "Swarm is encrypted and needs to be unlocked"
for _, d := range []*daemon.Daemon{d2, d3} {
assert.Equal(c, getNodeStatus(c, d), swarm.LocalNodeStateLocked)
outs, _ := d.Cmd("node", "ls")
assert.Assert(c, strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
cmd := d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(unlockKey)
result := icmd.RunCmd(cmd)
assert.Assert(c, strings.Contains(outs, swarmIsEncryptedMsg), outs)
// unlock with the original key should fail
// Use poll here because the daemon may not have finished
result := unlock(d, unlockKey)
if result.Error == nil {
// On occasion, the daemon may not have finished
// rotating the KEK before restarting. The test is
@ -1347,13 +1359,14 @@ func (s *DockerSwarmSuite) TestSwarmClusterRotateUnlockKey(c *testing.T) {
// restart again, the new key should be required this
// time.
time.Sleep(3 * time.Second)
// Wait for the rotation to happen
// Since there are multiple rotations, we need to wait until for the number of rotations we are currently on to be reflected in the logs
// This is a little janky... its depending on specific log messages AND these are debug logs... but it is the best we can do for now.
matcher := testdaemon.ScanLogsMatchCount(testdaemon.ScanLogsMatchString("successfully rotated KEK"), i+1)
poll.WaitOn(c, d.PollCheckLogs(ctx, matcher), poll.WithDelay(3*time.Second), poll.WithTimeout(time.Minute))
d.Restart(c)
d.RestartNode(c)
cmd = d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(unlockKey)
result = icmd.RunCmd(cmd)
result = unlock(d, unlockKey)
}
result.Assert(c, icmd.Expected{
ExitCode: 1,
@ -1361,31 +1374,21 @@ func (s *DockerSwarmSuite) TestSwarmClusterRotateUnlockKey(c *testing.T) {
})
outs, _ = d.Cmd("node", "ls")
assert.Assert(c, strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
cmd = d.Command("swarm", "unlock")
cmd.Stdin = bytes.NewBufferString(newUnlockKey)
icmd.RunCmd(cmd).Assert(c, icmd.Success)
assert.Assert(c, strings.Contains(outs, swarmIsEncryptedMsg), outs)
// now unlock with the rotated key, this should succeed
unlock(d, newUnlockKey).Assert(c, icmd.Success)
assert.Equal(c, getNodeStatus(c, d), swarm.LocalNodeStateActive)
retry := 0
for {
checkNodeLs := func(t poll.LogT) poll.Result {
// an issue sometimes prevents leader to be available right away
outs, err = d.Cmd("node", "ls")
if err != nil && retry < 5 {
if strings.Contains(outs, "swarm does not have a leader") {
retry++
c.Logf("[%s] got 'swarm does not have a leader'. retrying (attempt %d/5)", d.ID(), retry)
time.Sleep(3 * time.Second)
continue
} else {
c.Logf("[%s] gave error: '%v'. retrying (attempt %d/5): %s", d.ID(), err, retry, outs)
}
out, err := d.Cmd("node", "ls")
if err != nil {
return poll.Continue("error running node ls: %v: %s", err, out)
}
assert.NilError(c, err, "[%s] failed after %d retries: %v (%s)", d.ID(), retry, err, outs)
assert.Assert(c, !strings.Contains(outs, "Swarm is encrypted and needs to be unlocked"), outs)
break
return poll.Success()
}
poll.WaitOn(c, checkNodeLs, poll.WithDelay(3*time.Second), poll.WithTimeout(time.Minute))
}
unlockKey = newUnlockKey

11
testutil/daemon/daemon.go
View file

@ -335,6 +335,17 @@ func ScanLogsMatchString(contains string) func(string) bool {
}
}
// ScanLogsMatchCount returns a function that can be used to scan the daemon logs until the passed in matcher function matches `count` times
func ScanLogsMatchCount(f func(string) bool, count int) func(string) bool {
matched := 0
return func(line string) bool {
if f(line) {
matched++
}
return matched == count
}
}
// ScanLogsMatchAll returns a function that can be used to scan the daemon logs until *all* the passed in strings are matched
func ScanLogsMatchAll(contains ...string) func(string) bool {
matched := make(map[string]bool)