mirror of
https://github.com/hibiken/asynq.git
synced 2024-11-10 11:31:58 +08:00
989b2b6d55
Wait for a certain amount of time to allow for worker goroutines to finish. If the goroutines don't finish with the timeout duration, processor will quit the goroutines and restore any unfinished tasks from the in_progress queue back to the default queue.
117 lines
3.4 KiB
Go
117 lines
3.4 KiB
Go
package asynq
|
|
|
|
import (
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/go-redis/redis/v7"
|
|
"github.com/google/go-cmp/cmp"
|
|
"github.com/hibiken/asynq/internal/rdb"
|
|
)
|
|
|
|
func TestPoller(t *testing.T) {
|
|
type scheduledTask struct {
|
|
msg *rdb.TaskMessage
|
|
processAt time.Time
|
|
}
|
|
r := setup(t)
|
|
rdbClient := rdb.NewRDB(r)
|
|
const pollInterval = time.Second
|
|
p := newPoller(rdbClient, pollInterval)
|
|
t1 := randomTask("gen_thumbnail", "default", nil)
|
|
t2 := randomTask("send_email", "default", nil)
|
|
t3 := randomTask("reindex", "default", nil)
|
|
t4 := randomTask("sync", "default", nil)
|
|
|
|
tests := []struct {
|
|
initScheduled []scheduledTask // scheduled queue initial state
|
|
initRetry []scheduledTask // retry queue initial state
|
|
initQueue []*rdb.TaskMessage // default queue initial state
|
|
wait time.Duration // wait duration before checking for final state
|
|
wantScheduled []*rdb.TaskMessage // schedule queue final state
|
|
wantRetry []*rdb.TaskMessage // retry queue final state
|
|
wantQueue []*rdb.TaskMessage // default queue final state
|
|
}{
|
|
{
|
|
initScheduled: []scheduledTask{
|
|
{t1, time.Now().Add(time.Hour)},
|
|
{t2, time.Now().Add(-2 * time.Second)},
|
|
},
|
|
initRetry: []scheduledTask{
|
|
{t3, time.Now().Add(-500 * time.Millisecond)},
|
|
},
|
|
initQueue: []*rdb.TaskMessage{t4},
|
|
wait: pollInterval * 2,
|
|
wantScheduled: []*rdb.TaskMessage{t1},
|
|
wantRetry: []*rdb.TaskMessage{},
|
|
wantQueue: []*rdb.TaskMessage{t2, t3, t4},
|
|
},
|
|
{
|
|
initScheduled: []scheduledTask{
|
|
{t1, time.Now()},
|
|
{t2, time.Now().Add(-2 * time.Second)},
|
|
{t3, time.Now().Add(-500 * time.Millisecond)},
|
|
},
|
|
initRetry: []scheduledTask{},
|
|
initQueue: []*rdb.TaskMessage{t4},
|
|
wait: pollInterval * 2,
|
|
wantScheduled: []*rdb.TaskMessage{},
|
|
wantRetry: []*rdb.TaskMessage{},
|
|
wantQueue: []*rdb.TaskMessage{t1, t2, t3, t4},
|
|
},
|
|
}
|
|
|
|
for _, tc := range tests {
|
|
// clean up db before each test case.
|
|
if err := r.FlushDB().Err(); err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
// initialize scheduled queue
|
|
for _, st := range tc.initScheduled {
|
|
err := rdbClient.Schedule(st.msg, st.processAt)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
}
|
|
// initialize retry queue
|
|
for _, st := range tc.initRetry {
|
|
err := r.ZAdd(retryQ, &redis.Z{
|
|
Member: mustMarshal(t, st.msg),
|
|
Score: float64(st.processAt.Unix()),
|
|
}).Err()
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
}
|
|
// initialize default queue
|
|
for _, msg := range tc.initQueue {
|
|
err := rdbClient.Enqueue(msg)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
}
|
|
|
|
p.start()
|
|
time.Sleep(tc.wait)
|
|
p.terminate()
|
|
|
|
gotScheduledRaw := r.ZRange(scheduledQ, 0, -1).Val()
|
|
gotScheduled := mustUnmarshalSlice(t, gotScheduledRaw)
|
|
if diff := cmp.Diff(tc.wantScheduled, gotScheduled, sortMsgOpt); diff != "" {
|
|
t.Errorf("mismatch found in %q after running poller: (-want, +got)\n%s", scheduledQ, diff)
|
|
}
|
|
|
|
gotRetryRaw := r.ZRange(retryQ, 0, -1).Val()
|
|
gotRetry := mustUnmarshalSlice(t, gotRetryRaw)
|
|
if diff := cmp.Diff(tc.wantRetry, gotRetry, sortMsgOpt); diff != "" {
|
|
t.Errorf("mismatch found in %q after running poller: (-want, +got)\n%s", retryQ, diff)
|
|
}
|
|
|
|
gotQueueRaw := r.LRange(defaultQ, 0, -1).Val()
|
|
gotQueue := mustUnmarshalSlice(t, gotQueueRaw)
|
|
if diff := cmp.Diff(tc.wantQueue, gotQueue, sortMsgOpt); diff != "" {
|
|
t.Errorf("mismatch found in %q after running poller: (-want, +got)\n%s", defaultQ, diff)
|
|
}
|
|
}
|
|
}
|