Merge branch 'persistent_ha' into scheduler
This commit is contained in:
commit
a07a74a943
106 changed files with 18528 additions and 637 deletions
|
|
@ -76,6 +76,30 @@ var idCtxOnce sync.Once
|
|||
// Must be set via StartNotificationListener before RegisterToBackend.
|
||||
var gHandler NotificationHandler
|
||||
|
||||
func tryConnect(host string, port int, from, key string) error {
|
||||
cAddr := C.CString(host)
|
||||
cFrom := C.CString(from)
|
||||
cKey := C.CString(key)
|
||||
defer func() {
|
||||
C.free(unsafe.Pointer(cAddr))
|
||||
C.free(unsafe.Pointer(cFrom))
|
||||
C.free(unsafe.Pointer(cKey))
|
||||
}()
|
||||
|
||||
var cConfig C.xce_client_config_t
|
||||
cConfig.nmasters = 1
|
||||
cConfig.masters[0] = cAddr
|
||||
cConfig.master_ports[0] = C.int(port)
|
||||
cConfig.from = cFrom
|
||||
cConfig.key = cKey
|
||||
|
||||
rc := C.xce_backend_init(&cConfig, &gCtx)
|
||||
if rc < 0 {
|
||||
return fmt.Errorf("xce_backend_init failed on %s:%d: %d", host, port, int(rc))
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// RegisterToBackend initialises the XCE client library and joins as an API
|
||||
// server. StartNotificationListener should be called first so that push
|
||||
// events have a handler ready before the first push can arrive.
|
||||
|
|
@ -89,27 +113,31 @@ func RegisterToBackend(config *BackendConfig) error {
|
|||
key = "keepsecret$"
|
||||
}
|
||||
|
||||
cAddr := C.CString(config.Address)
|
||||
cFrom := C.CString(from)
|
||||
cKey := C.CString(key)
|
||||
defer func() {
|
||||
C.free(unsafe.Pointer(cAddr))
|
||||
C.free(unsafe.Pointer(cFrom))
|
||||
C.free(unsafe.Pointer(cKey))
|
||||
}()
|
||||
|
||||
var cConfig C.xce_client_config_t
|
||||
cConfig.nmasters = 1
|
||||
cConfig.masters[0] = cAddr
|
||||
cConfig.master_ports[0] = C.int(config.Port)
|
||||
cConfig.from = cFrom
|
||||
cConfig.key = cKey
|
||||
|
||||
rc := C.xce_backend_init(&cConfig, &gCtx)
|
||||
if rc < 0 {
|
||||
return fmt.Errorf("xce_backend_init failed: %d", int(rc))
|
||||
if len(config.Peers) > 0 {
|
||||
var lastErr error
|
||||
for _, peer := range config.Peers {
|
||||
host, port, err := parsePeer(peer)
|
||||
if err != nil {
|
||||
logger.Warn("Skipping invalid peer", zap.String("peer", peer), zap.Error(err))
|
||||
lastErr = err
|
||||
continue
|
||||
}
|
||||
if err := tryConnect(host, port, from, key); err != nil {
|
||||
logger.Warn("Candidate unreachable, trying next",
|
||||
zap.String("host", host), zap.Int("port", port), zap.Error(err))
|
||||
lastErr = err
|
||||
continue
|
||||
}
|
||||
logger.Info("Backend connection established (HA)",
|
||||
zap.String("host", host), zap.Int("port", port))
|
||||
return nil
|
||||
}
|
||||
return fmt.Errorf("all HA candidates failed: %w", lastErr)
|
||||
}
|
||||
|
||||
if err := tryConnect(config.Address, config.Port, from, key); err != nil {
|
||||
return err
|
||||
}
|
||||
logger.Info("Backend connection established",
|
||||
zap.String("address", config.Address),
|
||||
zap.Int("port", config.Port))
|
||||
|
|
|
|||
|
|
@ -7,4 +7,10 @@ type BackendConfig struct {
|
|||
Port int // controller port (e.g. 9990)
|
||||
From string // identity sent to controller (default: "xce_apiserver")
|
||||
Key string // authentication key
|
||||
|
||||
// Peers lists additional controller candidates for HA mode in "host:port"
|
||||
// format. When non-empty, RegisterToBackend tries each candidate in order
|
||||
// and connects to whichever accepts the join as primary. Address+Port
|
||||
// serve as a single-host fallback when Peers is empty.
|
||||
Peers []string
|
||||
}
|
||||
|
|
|
|||
|
|
@ -3,7 +3,6 @@ package backend
|
|||
import (
|
||||
"apiserver/models"
|
||||
"encoding/json"
|
||||
"time"
|
||||
)
|
||||
|
||||
// JobID bit layout constants — mirrors src/include/config.h.
|
||||
|
|
@ -179,18 +178,10 @@ func rawJobToModel(r rawJobData) *models.JobInfo {
|
|||
info.ExecUser = si.execUser
|
||||
}
|
||||
}
|
||||
if r.subTime > 0 {
|
||||
info.SubmitTime = time.Unix(r.subTime, 0)
|
||||
}
|
||||
if r.queueTime > 0 {
|
||||
info.QueueTime = time.Unix(r.queueTime, 0)
|
||||
}
|
||||
if r.startTime > 0 {
|
||||
info.StartTime = time.Unix(r.startTime, 0)
|
||||
}
|
||||
if r.finishTime > 0 {
|
||||
info.EndTime = time.Unix(r.finishTime, 0)
|
||||
}
|
||||
info.SubmitTime = r.subTime
|
||||
info.QueueTime = r.queueTime
|
||||
info.StartTime = r.startTime
|
||||
info.EndTime = r.finishTime
|
||||
if len(r.execHosts) > 0 {
|
||||
type hostSlot struct {
|
||||
HostName string `json:"host_name"`
|
||||
|
|
|
|||
|
|
@ -3,7 +3,6 @@ package backend
|
|||
import (
|
||||
"apiserver/models"
|
||||
"testing"
|
||||
"time"
|
||||
)
|
||||
|
||||
func TestRawJobToModel_BasicFields(t *testing.T) {
|
||||
|
|
@ -81,17 +80,17 @@ func TestRawJobToModel_Timestamps(t *testing.T) {
|
|||
|
||||
info := rawJobToModel(r)
|
||||
|
||||
if info.SubmitTime != time.Unix(1700000000, 0) {
|
||||
t.Errorf("SubmitTime mismatch")
|
||||
if info.SubmitTime != 1700000000 {
|
||||
t.Errorf("SubmitTime: got %d, want 1700000000", info.SubmitTime)
|
||||
}
|
||||
if info.QueueTime != time.Unix(1700000010, 0) {
|
||||
t.Errorf("QueueTime mismatch")
|
||||
if info.QueueTime != 1700000010 {
|
||||
t.Errorf("QueueTime: got %d, want 1700000010", info.QueueTime)
|
||||
}
|
||||
if info.StartTime != time.Unix(1700000020, 0) {
|
||||
t.Errorf("StartTime mismatch")
|
||||
if info.StartTime != 1700000020 {
|
||||
t.Errorf("StartTime: got %d, want 1700000020", info.StartTime)
|
||||
}
|
||||
if info.EndTime != time.Unix(1700000030, 0) {
|
||||
t.Errorf("EndTime mismatch")
|
||||
if info.EndTime != 1700000030 {
|
||||
t.Errorf("EndTime: got %d, want 1700000030", info.EndTime)
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
24
src/frontend/apiserver/backend/peers.go
Normal file
24
src/frontend/apiserver/backend/peers.go
Normal file
|
|
@ -0,0 +1,24 @@
|
|||
package backend
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"net"
|
||||
"strconv"
|
||||
)
|
||||
|
||||
// parsePeer splits a "host:port" string. Uses net.SplitHostPort so IPv6
|
||||
// bracket notation (e.g. "[::1]:9990") is handled correctly.
|
||||
func parsePeer(peer string) (host string, port int, err error) {
|
||||
h, portStr, err := net.SplitHostPort(peer)
|
||||
if err != nil {
|
||||
return "", 0, fmt.Errorf("invalid peer %q: %w", peer, err)
|
||||
}
|
||||
if h == "" {
|
||||
return "", 0, fmt.Errorf("invalid peer %q: empty host", peer)
|
||||
}
|
||||
port, err = strconv.Atoi(portStr)
|
||||
if err != nil || port <= 0 || port > 65535 {
|
||||
return "", 0, fmt.Errorf("invalid port in peer %q", peer)
|
||||
}
|
||||
return h, port, nil
|
||||
}
|
||||
95
src/frontend/apiserver/backend/peers_test.go
Normal file
95
src/frontend/apiserver/backend/peers_test.go
Normal file
|
|
@ -0,0 +1,95 @@
|
|||
package backend
|
||||
|
||||
import "testing"
|
||||
|
||||
func TestParsePeer(t *testing.T) {
|
||||
tests := []struct {
|
||||
name string
|
||||
peer string
|
||||
wantHost string
|
||||
wantPort int
|
||||
wantError bool
|
||||
}{
|
||||
{
|
||||
name: "valid simple host:port",
|
||||
peer: "ctrl01:9990",
|
||||
wantHost: "ctrl01",
|
||||
wantPort: 9990,
|
||||
},
|
||||
{
|
||||
name: "valid IPv4",
|
||||
peer: "192.168.1.10:9990",
|
||||
wantHost: "192.168.1.10",
|
||||
wantPort: 9990,
|
||||
},
|
||||
{
|
||||
name: "valid FQDN",
|
||||
peer: "ctrl01.cluster.local:9990",
|
||||
wantHost: "ctrl01.cluster.local",
|
||||
wantPort: 9990,
|
||||
},
|
||||
{
|
||||
name: "min valid port",
|
||||
peer: "host:1",
|
||||
wantHost: "host",
|
||||
wantPort: 1,
|
||||
},
|
||||
{
|
||||
name: "max valid port",
|
||||
peer: "host:65535",
|
||||
wantHost: "host",
|
||||
wantPort: 65535,
|
||||
},
|
||||
{
|
||||
name: "missing colon",
|
||||
peer: "ctrl019990",
|
||||
wantError: true,
|
||||
},
|
||||
{
|
||||
name: "port zero",
|
||||
peer: "ctrl01:0",
|
||||
wantError: true,
|
||||
},
|
||||
{
|
||||
name: "port out of range",
|
||||
peer: "ctrl01:65536",
|
||||
wantError: true,
|
||||
},
|
||||
{
|
||||
name: "non-numeric port",
|
||||
peer: "ctrl01:abc",
|
||||
wantError: true,
|
||||
},
|
||||
{
|
||||
name: "colon at start (empty host)",
|
||||
peer: ":9990",
|
||||
wantError: true,
|
||||
},
|
||||
{
|
||||
name: "empty string",
|
||||
peer: "",
|
||||
wantError: true,
|
||||
},
|
||||
}
|
||||
|
||||
for _, tc := range tests {
|
||||
t.Run(tc.name, func(t *testing.T) {
|
||||
host, port, err := parsePeer(tc.peer)
|
||||
if tc.wantError {
|
||||
if err == nil {
|
||||
t.Errorf("parsePeer(%q): expected error, got host=%q port=%d", tc.peer, host, port)
|
||||
}
|
||||
return
|
||||
}
|
||||
if err != nil {
|
||||
t.Fatalf("parsePeer(%q): unexpected error: %v", tc.peer, err)
|
||||
}
|
||||
if host != tc.wantHost {
|
||||
t.Errorf("host: got %q, want %q", host, tc.wantHost)
|
||||
}
|
||||
if port != tc.wantPort {
|
||||
t.Errorf("port: got %d, want %d", port, tc.wantPort)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
|
@ -29,4 +29,31 @@ func TestEvalSelectForHost_StubErrNoCGO(t *testing.T) {
|
|||
func TestCleanup_StubNoOp(t *testing.T) {
|
||||
// Should not panic and completes immediately
|
||||
Cleanup()
|
||||
}
|
||||
|
||||
// TestRegisterToBackend_StubSingleHost verifies the stub returns errNoCGO
|
||||
// for a single-host BackendConfig (no Peers).
|
||||
func TestRegisterToBackend_StubSingleHost(t *testing.T) {
|
||||
err := RegisterToBackend(&BackendConfig{
|
||||
Address: "ctrl01",
|
||||
Port: 9990,
|
||||
From: "xce_apiserver",
|
||||
Key: "keepsecret$",
|
||||
})
|
||||
if err != errNoCGO {
|
||||
t.Errorf("expected errNoCGO, got %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
// TestRegisterToBackend_StubWithPeers verifies the stub returns errNoCGO
|
||||
// even when Peers are configured (HA mode), since CGO is not available.
|
||||
func TestRegisterToBackend_StubWithPeers(t *testing.T) {
|
||||
err := RegisterToBackend(&BackendConfig{
|
||||
Peers: []string{"ctrl01:9990", "ctrl02:9990"},
|
||||
From: "xce_apiserver",
|
||||
Key: "keepsecret$",
|
||||
})
|
||||
if err != errNoCGO {
|
||||
t.Errorf("expected errNoCGO with peers configured, got %v", err)
|
||||
}
|
||||
}
|
||||
|
|
@ -66,6 +66,7 @@ func main() {
|
|||
Port: utils.BackendYamlConfig.Port,
|
||||
From: utils.BackendYamlConfig.From,
|
||||
Key: utils.BackendYamlConfig.Key,
|
||||
Peers: utils.BackendYamlConfig.Peers,
|
||||
}
|
||||
if err := backend.RegisterToBackend(bcfg); err != nil {
|
||||
logger.Error("Failed to connect to backend", zap.Error(err))
|
||||
|
|
|
|||
|
|
@ -1,7 +1,5 @@
|
|||
package models
|
||||
|
||||
import "time"
|
||||
|
||||
const (
|
||||
CONTROL_JOB = 1
|
||||
CONTROL_HOST = 2
|
||||
|
|
@ -11,9 +9,9 @@ const (
|
|||
)
|
||||
|
||||
type ActionControl struct {
|
||||
TargetType int `json:"target_type"`
|
||||
ActionType int `json:"action_type"`
|
||||
Content Jsonb `json:"content"`
|
||||
SubmitTime time.Time `json:"submit_time"`
|
||||
Submitter string `json:"submitter"`
|
||||
TargetType int `json:"target_type"`
|
||||
ActionType int `json:"action_type"`
|
||||
Content Jsonb `json:"content"`
|
||||
SubmitTime int64 `json:"submit_time"`
|
||||
Submitter string `json:"submitter"`
|
||||
}
|
||||
|
|
|
|||
|
|
@ -61,18 +61,19 @@ var JobSignalMapTable = map[string]int{
|
|||
}
|
||||
|
||||
var JOBSTAT = map[string]int{
|
||||
"PEND": 0x01,
|
||||
"RUN": 0x02,
|
||||
"EXIT": 0x04,
|
||||
"DONE": 0x08,
|
||||
"PSUSP": 0x10,
|
||||
"USUSP": 0x20,
|
||||
"SSUSP": 0x40,
|
||||
"PDONE": 0x80,
|
||||
"PERR": 0x100,
|
||||
"WAIT": 0x200,
|
||||
"RUNKWN": 0x8000,
|
||||
"UNKNOWN": 0x10000, // C JOB_STAT_UNKWN: worker lost contact with controller
|
||||
"PEND": 0x01,
|
||||
"RUN": 0x02,
|
||||
"EXIT": 0x04,
|
||||
"DONE": 0x08,
|
||||
"PSUSP": 0x10,
|
||||
"USUSP": 0x20,
|
||||
"SSUSP": 0x40,
|
||||
"PDONE": 0x80,
|
||||
"PERR": 0x100,
|
||||
"WAIT": 0x200,
|
||||
"TRANSIENT": 0x400, // C JOB_STAT_TRANSIENT: job state uncertain after failover
|
||||
"RUNKWN": 0x8000,
|
||||
"UNKNOWN": 0x10000, // C JOB_STAT_UNKWN: worker lost contact with controller
|
||||
}
|
||||
|
||||
var LAUNCHTYPE = map[string]int{
|
||||
|
|
@ -81,7 +82,10 @@ var LAUNCHTYPE = map[string]int{
|
|||
}
|
||||
|
||||
func JobStateStr(state int) string {
|
||||
if (state & JOBSTAT["PEND"]) != 0 {
|
||||
// Check TRANSIENT first — it takes priority after failover
|
||||
if (state & JOBSTAT["TRANSIENT"]) != 0 {
|
||||
return "TRANSIENT"
|
||||
} else if (state & JOBSTAT["PEND"]) != 0 {
|
||||
if (state & JOBSTAT["PSUSP"]) != 0 {
|
||||
return "PSUSP"
|
||||
}
|
||||
|
|
@ -485,7 +489,7 @@ type JobInfo struct {
|
|||
ArrayIdx int `json:"array_idx"` //new
|
||||
SubmitUser string `json:"submit_user"`
|
||||
FromHost string `json:"from_host"`
|
||||
SubmitTime time.Time `json:"submit_time"`
|
||||
SubmitTime int64 `json:"submit_time"`
|
||||
JobStatus int `json:"job_status"`
|
||||
JobName string `json:"job_name"`
|
||||
Command string `json:"command"`
|
||||
|
|
@ -531,10 +535,10 @@ type JobInfo struct {
|
|||
Tags Jsonb `json:"tags"` // new, JSONB as []byte
|
||||
EffectiveResReq string `json:"effective_resreq"` //new
|
||||
EffectiveUserGroup string `json:"effective_user_group"` //new
|
||||
QueueTime time.Time `json:"queue_time"`
|
||||
StartTime time.Time `json:"start_time"`
|
||||
EndTime time.Time `json:"end_time"`
|
||||
ReserveTime time.Time `json:"reserve_time"`
|
||||
QueueTime int64 `json:"queue_time"`
|
||||
StartTime int64 `json:"start_time"`
|
||||
EndTime int64 `json:"end_time"`
|
||||
ReserveTime int64 `json:"reserve_time"`
|
||||
Duration int `json:"duration"`
|
||||
CPUTime int `json:"cpu_time"`
|
||||
ExitStatus Jsonb `json:"exit_status"` //change type, JSONB as []byte
|
||||
|
|
@ -585,12 +589,12 @@ type JobAdditionalFields struct {
|
|||
JobID int64 `json:"job_id"`
|
||||
SubmitUser string `json:"submit_user"`
|
||||
JobStatus int `json:"job_status"`
|
||||
SubmitTime time.Time `json:"submit_time"`
|
||||
FromHost string `json:"from_host"`
|
||||
ExecHosts string `json:"exec_hosts"`
|
||||
SubmitTime int64 `json:"submit_time"`
|
||||
FromHost string `json:"from_host"`
|
||||
ExecHosts string `json:"exec_hosts"`
|
||||
|
||||
StartTime time.Time `json:"start_time"`
|
||||
EndTime time.Time `json:"end_time"`
|
||||
StartTime int64 `json:"start_time"`
|
||||
EndTime int64 `json:"end_time"`
|
||||
JobRusage Jsonb `json:"job_rusage"`
|
||||
MaxMem float64 `json:"max_mem"`
|
||||
AvgMem float64 `json:"avg_mem"`
|
||||
|
|
|
|||
|
|
@ -58,6 +58,16 @@ func TestJobStateStr(t *testing.T) {
|
|||
state: 0,
|
||||
want: "ERR",
|
||||
},
|
||||
{
|
||||
name: "TRANSIENT (0x400 — job uncertain after failover)",
|
||||
state: 0x400,
|
||||
want: "TRANSIENT",
|
||||
},
|
||||
{
|
||||
name: "TRANSIENT with other bits set",
|
||||
state: 0x400 | 0x01, // TRANSIENT + PEND
|
||||
want: "TRANSIENT",
|
||||
},
|
||||
}
|
||||
|
||||
for _, tt := range tests {
|
||||
|
|
@ -73,18 +83,19 @@ func TestJobStateStr(t *testing.T) {
|
|||
func TestJobStatMapValues(t *testing.T) {
|
||||
// Verify bitmask values match C JOB_STAT_* defines in src/include/xce/types.h.
|
||||
expected := map[string]int{
|
||||
"PEND": 0x01,
|
||||
"RUN": 0x02,
|
||||
"EXIT": 0x04,
|
||||
"DONE": 0x08,
|
||||
"PSUSP": 0x10,
|
||||
"USUSP": 0x20,
|
||||
"SSUSP": 0x40,
|
||||
"PDONE": 0x80,
|
||||
"PERR": 0x100,
|
||||
"WAIT": 0x200,
|
||||
"RUNKWN": 0x8000,
|
||||
"UNKNOWN": 0x10000,
|
||||
"PEND": 0x01,
|
||||
"RUN": 0x02,
|
||||
"EXIT": 0x04,
|
||||
"DONE": 0x08,
|
||||
"PSUSP": 0x10,
|
||||
"USUSP": 0x20,
|
||||
"SSUSP": 0x40,
|
||||
"PDONE": 0x80,
|
||||
"PERR": 0x100,
|
||||
"WAIT": 0x200,
|
||||
"TRANSIENT": 0x400,
|
||||
"RUNKWN": 0x8000,
|
||||
"UNKNOWN": 0x10000,
|
||||
}
|
||||
for key, want := range expected {
|
||||
got, ok := JOBSTAT[key]
|
||||
|
|
|
|||
|
|
@ -13,10 +13,15 @@ import (
|
|||
)
|
||||
|
||||
type BackendConfig struct {
|
||||
Address string `yaml:"address"`
|
||||
Port int `yaml:"port"`
|
||||
From string `yaml:"from"`
|
||||
Key string `yaml:"key"`
|
||||
Address string `yaml:"address"`
|
||||
Port int `yaml:"port"`
|
||||
From string `yaml:"from"`
|
||||
Key string `yaml:"key"`
|
||||
// Peers lists additional controller candidates for HA mode in "host:port"
|
||||
// format. When non-empty, RegisterToBackend tries each candidate in order
|
||||
// and connects to whichever accepts the join as primary. Address+Port
|
||||
// serve as a single-host fallback when Peers is empty.
|
||||
Peers []string `yaml:"peers"`
|
||||
}
|
||||
|
||||
type AppServerConfig struct {
|
||||
|
|
@ -250,6 +255,7 @@ func InitConf() error {
|
|||
Port: viper.GetInt("backend.port"),
|
||||
From: viper.GetString("backend.from"),
|
||||
Key: viper.GetString("backend.key"),
|
||||
Peers: viper.GetStringSlice("backend.peers"),
|
||||
}
|
||||
|
||||
return nil
|
||||
|
|
|
|||
|
|
@ -4,7 +4,6 @@ import (
|
|||
"encoding/json"
|
||||
"fmt"
|
||||
"strings"
|
||||
"time"
|
||||
jobfmt "xcecli/common/format"
|
||||
"xcecli/models"
|
||||
)
|
||||
|
|
@ -96,7 +95,7 @@ func (p *ArrayElementsPagePrinter) initColumnWidths(jobs *models.JobsResponse) {
|
|||
case models.HEADER_JOB_NAME:
|
||||
width = len(job.JobName)
|
||||
case models.HEADER_JOB_SUBMIT_TIME:
|
||||
width = len(job.SubmitTime)
|
||||
width = len(fmtUnixTime(job.SubmitTime))
|
||||
}
|
||||
|
||||
if !p.wideFormat && width > 15 {
|
||||
|
|
@ -166,7 +165,7 @@ func (p *ArrayElementsPagePrinter) formatRows(jobs *models.JobsResponse, truncat
|
|||
mainRow = appendOrDefault(mainRow, FormatStringField(job.FromHost, width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_EXEC_HOST:
|
||||
var execHosts models.ExecHosts
|
||||
if err := json.Unmarshal([]byte(job.ExecHosts.String()), &execHosts); err == nil && len(execHosts.HostSlots) > 0 {
|
||||
if err := json.Unmarshal([]byte(job.ExecHosts.String()), &execHosts); err == nil && len(execHosts.HostSlots) > 0 && execHosts.HostSlots[0].HostName != "" {
|
||||
firstHost := fmt.Sprintf("%d*%s", execHosts.HostSlots[0].NumSlots, execHosts.HostSlots[0].HostName)
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(firstHost, width, truncatFlag, true), "-")
|
||||
|
||||
|
|
@ -190,21 +189,7 @@ func (p *ArrayElementsPagePrinter) formatRows(jobs *models.JobsResponse, truncat
|
|||
case models.HEADER_JOB_NAME:
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(job.JobName, width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_SUBMIT_TIME:
|
||||
// remove the timezone part and parse the time
|
||||
timeStr := strings.Split(job.SubmitTime, " ")[0:2]
|
||||
timeStrWithoutTZ := strings.Join(timeStr, " ")
|
||||
|
||||
if submitTime, err := time.Parse("2006-01-02 15:04:05", timeStrWithoutTZ); err == nil {
|
||||
month := submitTime.Format("Jan")
|
||||
displayTime := fmt.Sprintf("%s %02d %02d:%02d",
|
||||
month,
|
||||
submitTime.Day(),
|
||||
submitTime.Hour(),
|
||||
submitTime.Minute())
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(displayTime, width, truncatFlag, true), "-")
|
||||
} else {
|
||||
mainRow = appendOrDefault(mainRow, fmt.Sprintf("%"+align+"*s", width, "-"), "-")
|
||||
}
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(fmtUnixTime(job.SubmitTime), width, truncatFlag, true), "-")
|
||||
}
|
||||
}
|
||||
rows = append(rows, mainRow)
|
||||
|
|
|
|||
|
|
@ -98,7 +98,7 @@ func (p *JobsPagePrinter) initColumnWidths(jobs *models.JobsResponse) {
|
|||
case models.HEADER_JOB_NAME:
|
||||
width = len(job.JobName)
|
||||
case models.HEADER_JOB_SUBMIT_TIME:
|
||||
width = len(job.SubmitTime)
|
||||
width = len(fmtUnixTime(job.SubmitTime))
|
||||
case models.HEADER_JOB_CORE_LIMIT:
|
||||
width = len(fmt.Sprintf("%f", job.CoreLimit))
|
||||
case models.HEADER_JOB_MEM_LIMIT:
|
||||
|
|
@ -179,7 +179,7 @@ func (p *JobsPagePrinter) formatRows(jobs *models.JobsResponse, truncatFlag bool
|
|||
|
||||
case models.HEADER_JOB_EXEC_HOST:
|
||||
var execHosts models.ExecHosts
|
||||
if err := json.Unmarshal([]byte(job.ExecHosts.String()), &execHosts); err == nil && len(execHosts.HostSlots) > 0 {
|
||||
if err := json.Unmarshal([]byte(job.ExecHosts.String()), &execHosts); err == nil && len(execHosts.HostSlots) > 0 && execHosts.HostSlots[0].HostName != "" {
|
||||
firstHost := fmt.Sprintf("%d*%s", execHosts.HostSlots[0].NumSlots, execHosts.HostSlots[0].HostName)
|
||||
mainRow = appendOrDefault(mainRow,
|
||||
FormatStringField(firstHost, width, truncatFlag, false), "-")
|
||||
|
|
@ -201,11 +201,7 @@ func (p *JobsPagePrinter) formatRows(jobs *models.JobsResponse, truncatFlag bool
|
|||
mainRow = appendOrDefault(mainRow, fmt.Sprintf("%"+align+"*s", width, "-"), "-")
|
||||
}
|
||||
case models.HEADER_JOB_SUBMIT_TIME:
|
||||
if submitTime, err := parseJobTime(job.SubmitTime); err == nil {
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(formatJobTime(submitTime), width, truncatFlag, true), "-")
|
||||
} else {
|
||||
mainRow = appendOrDefault(mainRow, fmt.Sprintf("%"+align+"*s", width, "-"), "-")
|
||||
}
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(fmtUnixTime(job.SubmitTime), width, truncatFlag, true), "-")
|
||||
|
||||
// normal fields
|
||||
case models.HEADER_JOB_NAME:
|
||||
|
|
@ -223,9 +219,9 @@ func (p *JobsPagePrinter) formatRows(jobs *models.JobsResponse, truncatFlag bool
|
|||
case models.HEADER_JOB_PROJECT_NAME:
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(job.ProjectName, width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_START_TIME:
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(job.StartTime, width, truncatFlag, true), "-")
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(fmtUnixTime(job.StartTime), width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_END_TIME:
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(job.EndTime, width, truncatFlag, true), "-")
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(fmtUnixTime(job.EndTime), width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_CWD:
|
||||
mainRow = appendOrDefault(mainRow, FormatStringField(job.CWD, width, truncatFlag, true), "-")
|
||||
case models.HEADER_JOB_LAUNCH_TYPE:
|
||||
|
|
@ -303,7 +299,7 @@ func printJobsVerboseOutput(jobs *models.JobsResponse, filter []string) {
|
|||
|
||||
// (2) print the submit time info
|
||||
submitLine := ""
|
||||
if submitTime, err := parseJobTime(job.SubmitTime); err == nil {
|
||||
if submitTime, ok := unixToTime(job.SubmitTime); ok {
|
||||
submitLine = submitLine + fmt.Sprintf("%s %s %02d:%02d:%02d: ",
|
||||
submitTime.Format("Jan"), submitTime.Format("02"),
|
||||
submitTime.Hour(), submitTime.Minute(), submitTime.Second())
|
||||
|
|
@ -322,9 +318,20 @@ func printJobsVerboseOutput(jobs *models.JobsResponse, filter []string) {
|
|||
}
|
||||
PrintWrappedLine(w, submitLine, 80, 18)
|
||||
|
||||
// (3) print the start time info
|
||||
// (3) print the pending reason for jobs waiting to be dispatched
|
||||
if job.JobStatus == "PEND" || job.JobStatus == "PSUSP" {
|
||||
if job.PendingReason != "" {
|
||||
fmt.Fprintf(w, " PENDING REASONS:\n")
|
||||
fmt.Fprintf(w, " %s\n", job.PendingReason)
|
||||
}
|
||||
}
|
||||
|
||||
// (4) print the start time info — skip for jobs that haven't started yet
|
||||
startLine := ""
|
||||
if startTime, err := parseJobTime(job.StartTime); err == nil {
|
||||
if job.JobStatus == "PEND" || job.JobStatus == "PSUSP" {
|
||||
goto printEndLine
|
||||
}
|
||||
if startTime, ok := unixToTime(job.StartTime); ok {
|
||||
month := startTime.Format("Jan")
|
||||
day := startTime.Format("02")
|
||||
exec_hostname := ""
|
||||
|
|
@ -380,10 +387,11 @@ func printJobsVerboseOutput(jobs *models.JobsResponse, filter []string) {
|
|||
}
|
||||
PrintWrappedLine(w, startLine, 80, 18)
|
||||
|
||||
printEndLine:
|
||||
// (4) print end line
|
||||
endLine := ""
|
||||
if job.JobStatus == "DONE" || job.JobStatus == "EXIT" {
|
||||
if endTime, err := parseJobTime(job.EndTime); err == nil {
|
||||
if endTime, ok := unixToTime(job.EndTime); ok {
|
||||
endLine = endLine + fmt.Sprintf("%s %s %02d:%02d:%02d: ",
|
||||
endTime.Format("Jan"), endTime.Format("02"),
|
||||
endTime.Hour(), endTime.Minute(), endTime.Second())
|
||||
|
|
|
|||
|
|
@ -5,17 +5,23 @@ import (
|
|||
"time"
|
||||
)
|
||||
|
||||
// parseJobTime parses a job time string supporting RFC3339 and space-separated formats.
|
||||
func parseJobTime(s string) (time.Time, error) {
|
||||
if s == "" {
|
||||
return time.Time{}, fmt.Errorf("empty time string")
|
||||
// unixToTime converts a Unix-second timestamp to time.Time.
|
||||
// Returns (time, true) for ts > 0; (zero, false) for unset/zero values.
|
||||
func unixToTime(ts int64) (time.Time, bool) {
|
||||
if ts <= 0 {
|
||||
return time.Time{}, false
|
||||
}
|
||||
for _, f := range []string{time.RFC3339Nano, time.RFC3339, "2006-01-02 15:04:05", "2006-01-02 15:04:05 MST"} {
|
||||
if t, err := time.Parse(f, s); err == nil {
|
||||
return t, nil
|
||||
}
|
||||
return time.Unix(ts, 0), true
|
||||
}
|
||||
|
||||
// fmtUnixTime formats a Unix timestamp as "Jan 02 15:04" for table display.
|
||||
// Returns "-" for zero/unset values.
|
||||
func fmtUnixTime(ts int64) string {
|
||||
t, ok := unixToTime(ts)
|
||||
if !ok {
|
||||
return "-"
|
||||
}
|
||||
return time.Time{}, fmt.Errorf("cannot parse time: %q", s)
|
||||
return formatJobTime(t)
|
||||
}
|
||||
|
||||
// formatJobTime formats a time.Time as "Jan 02 15:04" for table display.
|
||||
|
|
|
|||
|
|
@ -3,8 +3,6 @@ package models
|
|||
import (
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"time"
|
||||
|
||||
)
|
||||
|
||||
const (
|
||||
|
|
@ -32,6 +30,7 @@ const (
|
|||
HEADER_JOB_LAUNCH_TYPE = "LAUNCH_TYPE"
|
||||
HEADER_JOB_NUM_PROCESSORS = "NUM_PROCESSORS"
|
||||
HEADER_JOB_INTERACTIVE_URL = "INTERACTIVE_URL"
|
||||
HEADER_JOB_PENDING_REASON = "PENDING_REASON"
|
||||
|
||||
// second line
|
||||
HEADER_JOB_R15S = "r15s"
|
||||
|
|
@ -114,10 +113,10 @@ type QueryJob struct {
|
|||
TermTime string `json:"term_time"`
|
||||
UserPriority int `json:"user_priority"`
|
||||
DependCond string `json:"depend_cond"`
|
||||
SubmitTime string `json:"submit_time"`
|
||||
ReserveTime time.Time `json:"reserve_time"`
|
||||
StartTime string `json:"start_time"`
|
||||
EndTime string `json:"end_time"`
|
||||
SubmitTime int64 `json:"submit_time"`
|
||||
ReserveTime int64 `json:"reserve_time"`
|
||||
StartTime int64 `json:"start_time"`
|
||||
EndTime int64 `json:"end_time"`
|
||||
JobRusage string `json:"job_rusage"`
|
||||
MaxMem float64 `json:"max_mem"`
|
||||
AvgMem float64 `json:"avg_mem"`
|
||||
|
|
@ -210,6 +209,7 @@ var JobFieldsMap = map[string]string{
|
|||
HEADER_JOB_LAUNCH_TYPE: "launch_type",
|
||||
HEADER_JOB_NUM_PROCESSORS: "num_processors",
|
||||
HEADER_JOB_INTERACTIVE_URL: "interactive_job_url",
|
||||
HEADER_JOB_PENDING_REASON: "pending_reason",
|
||||
//job_filter.HEADER_JOB_PRIORITY: "job_priority",
|
||||
}
|
||||
|
||||
|
|
@ -258,16 +258,23 @@ func (e ExecHosts) String() string {
|
|||
|
||||
// UnmarshalJSON customizes the JSON parsing method, compatible with both string "-" and structure formats
|
||||
func (e *ExecHosts) UnmarshalJSON(data []byte) error {
|
||||
// JSON null means no allocation — treat as empty
|
||||
if string(data) == "null" {
|
||||
e.HostSlots = []HostSlot{}
|
||||
e.NumHosts = 0
|
||||
return nil
|
||||
}
|
||||
|
||||
// First try to parse as a string
|
||||
var str string
|
||||
if err := json.Unmarshal(data, &str); err == nil {
|
||||
// If the string is "-", return an empty ExecHosts
|
||||
if str == "-" {
|
||||
// Empty string or "-" both mean no hosts
|
||||
if str == "" || str == "-" {
|
||||
e.HostSlots = []HostSlot{}
|
||||
e.NumHosts = 0
|
||||
return nil
|
||||
}
|
||||
// If it's any other string, treat it as a single hostname
|
||||
// Any other non-empty string is a single hostname
|
||||
e.HostSlots = []HostSlot{{HostName: str, NumSlots: 1}}
|
||||
e.NumHosts = 1
|
||||
return nil
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue