Fix timing issue in sentinel CKQUORUM test (#10036)

A test failure was reported in Daily CI (test-centos7-tls).
`CKQUORUM detects failover authorization cannot be reached`.

```
CKQUORUM detects failover authorization cannot be reached: FAILED:
Expected 'invalid command name "OK 4 usable Sentinels. Quorum and failover authorization can be reached"' to match '*NOQUORUM*'
```

It seems that current sentinel does not confirm that the other
sentinels are actually `down`, and then check the quorum.
It at least take 3 seconds on my machine, and we can see there
will be a timing issue with the hard code `after 5000`.

In this commit, we check the response of `SENTINEL SENTINELS mymaster`
to ensure that other sentinels are actually `down` in the view the
current sentinel. Solve the timing issue due to sentinel monitor mechanism.

1 files changed, 10 insertions, 3 deletions

diff --git a/tests/sentinel/tests/06-ckquorum.tcl b/tests/sentinel/tests/06-ckquorum.tcl
index 4ea4e55d8..36c3dc650 100644
--- a/tests/sentinel/tests/06-ckquorum.tcl
+++ b/tests/sentinel/tests/06-ckquorum.tcl
@@ -24,9 +24,16 @@ test "CKQUORUM detects failover authorization cannot be reached" {
         kill_instance sentinel [expr {$i + 1}]
     }
 
-    after 5000
-    catch {[S 0 SENTINEL CKQUORUM mymaster]} err
-    assert_match "*NOQUORUM*" $err
+    # We need to make sure that other sentinels are in `DOWN` state
+    # from the point of view of S 0 before we executing `CKQUORUM`.
+    wait_for_condition 300 50 {
+        [catch {S 0 SENTINEL CKQUORUM mymaster}] == 1
+    } else {
+        fail "At least $orig_quorum sentinels did not enter the down state."
+    }
+
+    assert_error "*NOQUORUM*" {S 0 SENTINEL CKQUORUM mymaster}
+
     S 0 SENTINEL SET mymaster quorum $orig_quorum
     for {set i 0} {$i < $orig_quorum} {incr i} {
         restart_instance sentinel [expr {$i + 1}]