// checkGossip fetches the gossip infoStore from each node and invokes the given
// function. The test passes if the function returns 0 for every node,
// retrying for up to the given duration.
func checkGossip(t *testing.T, c cluster.Cluster, d time.Duration,
f checkGossipFunc) {
util.SucceedsWithin(t, d, func() error {
select {
case <-stopper:
t.Fatalf("interrupted")
return nil
case <-time.After(1 * time.Second):
}
for i := 0; i < c.NumNodes(); i++ {
var m map[string]interface{}
if err := getJSON(c.URL(i), "/_status/gossip/local", &m); err != nil {
return err
}
infos, ok := m["infos"].(map[string]interface{})
if !ok {
return errors.New("no infos yet")
}
if err := f(infos); err != nil {
return util.Errorf("node %d: %s", i, err)
}
}
return nil
})
}
// checkGossip fetches the gossip infoStore from each node and invokes the given
// function. The test passes if the function returns 0 for every node,
// retrying for up to the given duration.
func checkGossip(t *testing.T, c cluster.Cluster, d time.Duration, f checkGossipFunc) {
err := util.RetryForDuration(d, func() error {
select {
case <-stopper:
t.Fatalf("interrupted")
return nil
case <-time.After(1 * time.Second):
}
var infoStatus gossip.InfoStatus
for i := 0; i < c.NumNodes(); i++ {
if err := util.GetJSON(cluster.HTTPClient, c.URL(i)+"/_status/gossip/local", &infoStatus); err != nil {
return err
}
if err := f(infoStatus.Infos); err != nil {
return errors.Errorf("node %d: %s", i, err)
}
}
return nil
})
if err != nil {
t.Fatal(errors.Errorf("condition failed to evaluate within %s: %s", d, err))
}
}
func testBuildInfoInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
checkGossip(t, c, 20*time.Second, hasPeers(c.NumNodes()))
var details server.DetailsResponse
util.SucceedsSoon(t, func() error {
select {
case <-stopper:
t.Fatalf("interrupted")
default:
}
return util.GetJSON(cluster.HTTPClient, c.URL(0)+"/_status/details/local", &details)
})
bi := details.BuildInfo
testData := map[string]string{
"go_version": bi.GoVersion,
"tag": bi.Tag,
"time": bi.Time,
"dependencies": bi.Dependencies,
}
for key, val := range testData {
if val == "" {
t.Errorf("build info not set for \"%s\"", key)
}
}
}
func testStatusServerInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
// Get the ids for each node.
idMap := make(map[int]string)
for i := 0; i < c.NumNodes(); i++ {
var detail details
if err := getJSON(c.URL(i), "/_status/details/local", &detail); err != nil {
t.Fatal(err)
}
idMap[i] = detail.NodeID.String()
}
// Check local response for the every node.
for i := 0; i < c.NumNodes(); i++ {
checkNode(t, c, i, idMap[i], "local", idMap[i])
get(t, c.URL(i), "/_status/nodes")
get(t, c.URL(i), "/_status/stores")
}
// Proxy from the first node to the last node.
firstNode := 0
lastNode := c.NumNodes() - 1
firstID := idMap[firstNode]
lastID := idMap[lastNode]
checkNode(t, c, firstNode, firstID, lastID, lastID)
// And from the last node to the first node.
checkNode(t, c, lastNode, lastID, firstID, firstID)
// And from the last node to the last node.
checkNode(t, c, lastNode, lastID, lastID, lastID)
}
func postFreeze(c cluster.Cluster, freeze bool, timeout time.Duration) (serverpb.ClusterFreezeResponse, error) {
httpClient := cluster.HTTPClient
httpClient.Timeout = timeout
var resp serverpb.ClusterFreezeResponse
log.Infof("requesting: freeze=%t, timeout=%s", freeze, timeout)
cb := func(v proto.Message) {
oldNum := resp.RangesAffected
resp = *v.(*serverpb.ClusterFreezeResponse)
if oldNum > resp.RangesAffected {
resp.RangesAffected = oldNum
}
if (resp != serverpb.ClusterFreezeResponse{}) {
log.Infof("%+v", &resp)
}
}
err := util.StreamJSON(
httpClient,
c.URL(0)+"/_admin/v1/cluster/freeze",
&serverpb.ClusterFreezeRequest{Freeze: freeze},
&serverpb.ClusterFreezeResponse{},
cb,
)
return resp, err
}
func testStatusServerInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
// Get the ids for each node.
idMap := make(map[int]roachpb.NodeID)
for i := 0; i < c.NumNodes(); i++ {
var details server.DetailsResponse
if err := util.GetJSON(cluster.HTTPClient, c.URL(i)+"/_status/details/local", &details); err != nil {
t.Fatal(err)
}
idMap[i] = details.NodeID
}
// Check local response for the every node.
for i := 0; i < c.NumNodes(); i++ {
id := idMap[i]
checkNode(t, c, i, id, id, id)
get(t, c.URL(i), "/_status/nodes")
}
// Proxy from the first node to the last node.
firstNode := 0
lastNode := c.NumNodes() - 1
firstID := idMap[firstNode]
lastID := idMap[lastNode]
checkNode(t, c, firstNode, firstID, lastID, lastID)
// And from the last node to the first node.
checkNode(t, c, lastNode, lastID, firstID, firstID)
// And from the last node to the last node.
checkNode(t, c, lastNode, lastID, lastID, lastID)
}
func testBuildInfoInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
checkGossip(t, c, 20*time.Second, hasPeers(c.NumNodes()))
util.SucceedsSoon(t, func() error {
select {
case <-stopper:
t.Fatalf("interrupted")
return nil
default:
}
var r struct {
BuildInfo map[string]string
}
if err := getJSON(c.URL(0), "/_status/details/local", &r); err != nil {
return err
}
for _, key := range []string{"goVersion", "tag", "time", "dependencies"} {
if val, ok := r.BuildInfo[key]; !ok {
t.Errorf("build info missing for \"%s\"", key)
} else if val == "" {
t.Errorf("build info not set for \"%s\"", key)
}
}
return nil
})
}
func postFreeze(c cluster.Cluster, freeze bool) (server.ClusterFreezeResponse, error) {
httpClient := cluster.HTTPClient()
httpClient.Timeout = 10 * time.Second
var resp server.ClusterFreezeResponse
err := postJSON(httpClient, c.URL(0), "/_admin/v1/cluster/freeze",
&server.ClusterFreezeRequest{Freeze: freeze}, &resp)
return resp, err
}
func postFreeze(c cluster.Cluster, freeze bool, timeout time.Duration) (server.ClusterFreezeResponse, error) {
httpClient := cluster.HTTPClient
httpClient.Timeout = timeout
var resp server.ClusterFreezeResponse
err := util.PostJSON(
httpClient,
c.URL(0)+"/_admin/v1/cluster/freeze",
&server.ClusterFreezeRequest{Freeze: freeze},
&resp,
)
return resp, err
}
func testAdminLossOfQuorumInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
if c.NumNodes() < 2 {
t.Logf("skipping test %s because given cluster has too few nodes", cfg.Name)
return
}
// Get the ids for each node.
nodeIDs := make([]roachpb.NodeID, c.NumNodes())
for i := 0; i < c.NumNodes(); i++ {
var details serverpb.DetailsResponse
if err := util.GetJSON(cluster.HTTPClient, c.URL(i)+"/_status/details/local", &details); err != nil {
t.Fatal(err)
}
nodeIDs[i] = details.NodeID
}
// Leave only the first node alive.
for i := 1; i < c.NumNodes(); i++ {
if err := c.Kill(i); err != nil {
t.Fatal(err)
}
}
// Retrieve node statuses.
var nodes serverpb.NodesResponse
if err := util.GetJSON(cluster.HTTPClient, c.URL(0)+"/_status/nodes", &nodes); err != nil {
t.Fatal(err)
}
for _, nodeID := range nodeIDs {
var nodeStatus status.NodeStatus
if err := util.GetJSON(cluster.HTTPClient, c.URL(0)+"/_status/nodes/"+strconv.Itoa(int(nodeID)), &nodeStatus); err != nil {
t.Fatal(err)
}
}
// Retrieve time-series data.
nowNanos := timeutil.Now().UnixNano()
queryRequest := tspb.TimeSeriesQueryRequest{
StartNanos: nowNanos - 10*time.Second.Nanoseconds(),
EndNanos: nowNanos,
Queries: []tspb.Query{
{Name: "doesnt_matter", Sources: []string{}},
},
}
var queryResponse tspb.TimeSeriesQueryResponse
if err := util.PostJSON(cluster.HTTPClient, c.URL(0)+"/ts/query",
&queryRequest, &queryResponse); err != nil {
t.Fatal(err)
}
// TODO(cdo): When we're able to issue SQL queries without a quorum, test all
// admin endpoints that issue SQL queries here.
}
func testAdminLossOfQuorumInner(t *testing.T, c cluster.Cluster, cfg cluster.TestConfig) {
if c.NumNodes() < 2 {
t.Logf("skipping test %s because given cluster has too few nodes", cfg.Name)
return
}
// Get the ids for each node.
idMap := make(map[int]string)
for i := 0; i < c.NumNodes(); i++ {
var detail details
if err := getJSON(c.URL(i), "/_status/details/local", &detail); err != nil {
t.Fatal(err)
}
idMap[i] = detail.NodeID.String()
}
// Leave only the first node alive.
for i := 1; i < c.NumNodes(); i++ {
if err := c.Kill(i); err != nil {
t.Fatal(err)
}
}
// Retrieve node statuses.
var nodeStatuses interface{}
if err := getJSON(c.URL(0), "/_status/nodes/", &nodeStatuses); err != nil {
t.Fatal(err)
}
for i := 0; i < c.NumNodes(); i++ {
var nodeStatus interface{}
url := fmt.Sprintf("/_status/nodes/%s", idMap[i])
if err := getJSON(c.URL(0), url, &nodeStatus); err != nil {
t.Fatal(err)
}
}
// Retrieve time-series data.
nowNanos := timeutil.Now().UnixNano()
queryRequest := ts.TimeSeriesQueryRequest{
StartNanos: nowNanos - 10*time.Second.Nanoseconds(),
EndNanos: nowNanos,
Queries: []ts.Query{
{Name: "doesnt_matter", Sources: []string{}},
},
}
var queryResponse ts.TimeSeriesQueryResponse
if err := postJSON(cluster.HTTPClient(), c.URL(0), "/ts/query",
&queryRequest, &queryResponse); err != nil {
t.Fatal(err)
}
// TODO(cdo): When we're able to issue SQL queries without a quorum, test all
// admin endpoints that issue SQL queries here.
}
// checkNode checks all the endpoints of the status server hosted by node and
// requests info for the node with otherNodeID. That node could be the same
// other node, the same node or "local".
func checkNode(t *testing.T, c cluster.Cluster, i int, nodeID, otherNodeID, expectedNodeID string) {
var detail details
if err := getJSON(c.URL(i), "/_status/details/"+otherNodeID, &detail); err != nil {
t.Fatal(util.ErrorfSkipFrames(1, "unable to parse details - %s", err))
}
if actualNodeID := detail.NodeID.String(); actualNodeID != expectedNodeID {
t.Fatal(util.ErrorfSkipFrames(1, "%s calling %s: node ids don't match - expected %s, actual %s", nodeID, otherNodeID, expectedNodeID, actualNodeID))
}
get(t, c.URL(i), fmt.Sprintf("/_status/gossip/%s", otherNodeID))
get(t, c.URL(i), fmt.Sprintf("/_status/logfiles/%s", otherNodeID))
get(t, c.URL(i), fmt.Sprintf("/_status/logs/%s", otherNodeID))
get(t, c.URL(i), fmt.Sprintf("/_status/stacks/%s", otherNodeID))
get(t, c.URL(i), fmt.Sprintf("/_status/nodes/%s", otherNodeID))
}
// checkNode checks all the endpoints of the status server hosted by node and
// requests info for the node with otherNodeID. That node could be the same
// other node, the same node or "local".
func checkNode(t *testing.T, c cluster.Cluster, i int, nodeID, otherNodeID, expectedNodeID roachpb.NodeID) {
urlIDs := []string{otherNodeID.String()}
if nodeID == otherNodeID {
urlIDs = append(urlIDs, "local")
}
var details server.DetailsResponse
for _, urlID := range urlIDs {
if err := util.GetJSON(cluster.HTTPClient, c.URL(i)+"/_status/details/"+urlID, &details); err != nil {
t.Fatal(util.ErrorfSkipFrames(1, "unable to parse details - %s", err))
}
if details.NodeID != expectedNodeID {
t.Fatal(util.ErrorfSkipFrames(1, "%d calling %s: node ids don't match - expected %d, actual %d", nodeID, urlID, expectedNodeID, details.NodeID))
}
get(t, c.URL(i), fmt.Sprintf("/_status/gossip/%s", urlID))
get(t, c.URL(i), fmt.Sprintf("/_status/nodes/%s", urlID))
get(t, c.URL(i), fmt.Sprintf("/_status/logfiles/%s", urlID))
get(t, c.URL(i), fmt.Sprintf("/_status/logs/%s", urlID))
get(t, c.URL(i), fmt.Sprintf("/_status/stacks/%s", urlID))
}
}