-
Notifications
You must be signed in to change notification settings - Fork 21
/
task.go
307 lines (281 loc) · 8.69 KB
/
task.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
package core
import (
"errors"
"fmt"
"github.com/MG-RAST/AWE/lib/conf"
"github.com/MG-RAST/AWE/lib/logger"
"os/exec"
"strconv"
"strings"
"time"
)
const (
TASK_STAT_INIT = "init"
TASK_STAT_QUEUED = "queued"
TASK_STAT_INPROGRESS = "in-progress"
TASK_STAT_PENDING = "pending"
TASK_STAT_SUSPEND = "suspend"
TASK_STAT_COMPLETED = "completed"
TASK_STAT_SKIPPED = "user_skipped"
TASK_STAT_FAIL_SKIP = "skipped"
TASK_STAT_PASSED = "passed"
)
type Task struct {
Id string `bson:"taskid" json:"taskid"`
Info *Info `bson:"info" json:"-"`
Inputs IOmap `bson:"inputs" json:"inputs"`
Outputs IOmap `bson:"outputs" json:"outputs"`
Predata IOmap `bson:"predata" json:"predata"`
Cmd *Command `bson:"cmd" json:"cmd"`
Partition *PartInfo `bson:"partinfo" json:"-"`
DependsOn []string `bson:"dependsOn" json:"dependsOn"`
TotalWork int `bson:"totalwork" json:"totalwork"`
MaxWorkSize int `bson:"maxworksize" json:"maxworksize"`
RemainWork int `bson:"remainwork" json:"remainwork"`
WorkStatus []string `bson:"workstatus" json:"-"`
State string `bson:"state" json:"state"`
Skip int `bson:"skip" json:"-"`
CreatedDate time.Time `bson:"createdDate" json:"createddate"`
StartedDate time.Time `bson:"startedDate" json:"starteddate"`
CompletedDate time.Time `bson:"completedDate" json:"completeddate"`
ComputeTime int `bson:"computetime" json:"computetime"`
UserAttr map[string]string `bson:"userattr" json:"userattr"`
AppVariables AppVariables
}
func NewTask(job *Job, rank int) *Task {
return &Task{
Id: fmt.Sprintf("%s_%d", job.Id, rank),
Info: job.Info,
Inputs: NewIOmap(),
Outputs: NewIOmap(),
Cmd: &Command{},
Partition: nil,
DependsOn: []string{},
TotalWork: 1,
RemainWork: 1,
WorkStatus: []string{},
State: TASK_STAT_INIT,
Skip: 0,
}
}
// fill some info (lacked in input json) for a task
func (task *Task) InitTask(job *Job, rank int) (err error) {
//validate taskid
if len(task.Id) == 0 {
return errors.New("invalid taskid:" + task.Id)
}
parts := strings.Split(task.Id, "_")
if len(parts) == 2 {
//is standard taskid (%s_%d), do nothing
} else if idInt, err := strconv.Atoi(task.Id); err == nil {
//if task.Id is an "integer", it is unmashalled from job.json (submitted by template)
//convert to standard taskid
if rank != idInt {
return errors.New(fmt.Sprintf("invalid job script: task id doen't match stage %d vs %d", rank, idInt))
}
task.Id = fmt.Sprintf("%s_%s", job.Id, task.Id)
for j := 0; j < len(task.DependsOn); j++ {
depend := task.DependsOn[j]
task.DependsOn[j] = fmt.Sprintf("%s_%s", job.Id, depend)
}
} else {
return errors.New("invalid taskid:" + task.Id)
}
task.Info = job.Info
if task.TotalWork <= 0 {
task.setTotalWork(1)
}
task.WorkStatus = make([]string, task.TotalWork)
task.RemainWork = task.TotalWork
for _, io := range task.Inputs {
if io.Node == "" {
io.Node = "-"
}
}
for _, io := range task.Outputs {
if io.Node == "" {
io.Node = "-"
}
}
if len(task.Cmd.Environ.Private) > 0 {
task.Cmd.HasPrivateEnv = true
}
task.setTokenForIO()
task.State = TASK_STAT_INIT
return
}
func (task *Task) UpdateState(newState string) string {
task.State = newState
return task.State
}
func (task *Task) CreateIndex() (err error) {
for _, io := range task.Inputs {
if len(io.ShockIndex) > 0 {
idxinfo, err := io.GetIndexInfo()
if err != nil {
errMsg := "could not retrieve index info from input shock node, taskid=" + task.Id
logger.Error("error: " + errMsg)
return errors.New(errMsg)
}
if _, ok := idxinfo[io.ShockIndex]; !ok {
if err := ShockPutIndex(io.Host, io.Node, io.ShockIndex, task.Info.DataToken); err != nil {
errMsg := "failed to create index on shock node for taskid=" + task.Id
logger.Error("error: " + errMsg)
return errors.New(errMsg)
}
}
}
}
return
}
//get part size based on partition/index info
//if fail to get index info, task.TotalWork fall back to 1 and return nil
func (task *Task) InitPartIndex() (err error) {
if task.TotalWork == 1 && task.MaxWorkSize == 0 {
return
}
var input_io *IO
if task.Partition == nil {
if len(task.Inputs) == 1 {
for filename, io := range task.Inputs {
input_io = io
task.Partition = new(PartInfo)
task.Partition.Input = filename
task.Partition.MaxPartSizeMB = task.MaxWorkSize
break
}
} else {
task.setTotalWork(1)
logger.Error("warning: lacking parition info while multiple inputs are specified, taskid=" + task.Id)
return
}
} else {
if task.MaxWorkSize > 0 {
task.Partition.MaxPartSizeMB = task.MaxWorkSize
}
if task.Partition.MaxPartSizeMB == 0 && task.TotalWork <= 1 {
task.setTotalWork(1)
return
}
if _, ok := task.Inputs[task.Partition.Input]; !ok {
task.setTotalWork(1)
logger.Error("warning: invalid partition info, taskid=" + task.Id)
return
}
input_io = task.Inputs[task.Partition.Input]
}
var totalunits int
idxinfo, err := input_io.GetIndexInfo()
if err != nil {
task.setTotalWork(1)
logger.Error("warning: invalid file info, taskid=" + task.Id)
return nil
}
idxtype := conf.DEFAULT_INDEX
if _, ok := idxinfo[idxtype]; !ok { //if index not available, create index
if err := ShockPutIndex(input_io.Host, input_io.Node, idxtype, task.Info.DataToken); err != nil {
task.setTotalWork(1)
logger.Error("warning: fail to create index on shock for taskid=" + task.Id)
return nil
}
totalunits, err = input_io.TotalUnits(idxtype) //get index info again
if err != nil {
task.setTotalWork(1)
logger.Error("warning: fail to get index units, taskid=" + task.Id + ":" + err.Error())
return nil
}
} else { //index existing, use it directly
totalunits = int(idxinfo[idxtype].TotalUnits)
}
//adjust total work based on needs
if task.Partition.MaxPartSizeMB > 0 { // fixed max part size
//this implementation for chunkrecord indexer only
chunkmb := int(conf.DEFAULT_CHUNK_SIZE / 1048576)
var totalwork int
if totalunits*chunkmb%task.Partition.MaxPartSizeMB == 0 {
totalwork = totalunits * chunkmb / task.Partition.MaxPartSizeMB
} else {
totalwork = totalunits*chunkmb/task.Partition.MaxPartSizeMB + 1
}
if totalwork < task.TotalWork { //use bigger splits (specified by size or totalwork)
totalwork = task.TotalWork
}
task.setTotalWork(totalwork)
}
if totalunits < task.TotalWork {
task.setTotalWork(totalunits)
}
task.Partition.Index = idxtype
task.Partition.TotalIndex = totalunits
return
}
func (task *Task) setTotalWork(num int) {
task.TotalWork = num
task.RemainWork = num
task.WorkStatus = make([]string, num)
}
func (task *Task) setTokenForIO() {
if !task.Info.Auth || task.Info.DataToken == "" {
return
}
for _, io := range task.Inputs {
io.DataToken = task.Info.DataToken
}
for _, io := range task.Outputs {
io.DataToken = task.Info.DataToken
}
}
func (task *Task) ParseWorkunit() (wus []*Workunit, err error) {
//if a task contains only one workunit, assign rank 0
if task.TotalWork == 1 {
workunit := NewWorkunit(task, 0)
wus = append(wus, workunit)
return
}
// if a task contains N (N>1) workunits, assign rank 1..N
for i := 1; i <= task.TotalWork; i++ {
workunit := NewWorkunit(task, i)
wus = append(wus, workunit)
}
return
}
func (task *Task) Skippable() bool {
// For a task to be skippable, it should meet
// the following requirements (this may change
// in the future):
// 1.- It should have exactly one input file
// and one output file (This way, we can connect tasks
// Ti-1 and Ti+1 transparently)
// 2.- It should be a simple pipeline task. That is,
// it should just have at most one "parent" Ti-1 ---> Ti
return (len(task.Inputs) == 1) &&
(len(task.Outputs) == 1) &&
(len(task.DependsOn) <= 1)
}
func (task *Task) DeleteOutput() {
if task.State == TASK_STAT_COMPLETED ||
task.State == TASK_STAT_SKIPPED ||
task.State == TASK_STAT_FAIL_SKIP {
for _, io := range task.Outputs {
if io.Delete {
if nodeid, err := io.DeleteNode(); err != nil {
logger.Error(fmt.Sprintf("warning: fail to delete shock node %s: %s", nodeid, err.Error()))
}
}
}
}
}
//creat index (=deprecated=)
func createIndex(host string, nodeid string, indexname string) (err error) {
argv := []string{}
argv = append(argv, "-X")
argv = append(argv, "PUT")
target_url := fmt.Sprintf("%s/node/%s?index=%s", host, nodeid, indexname)
argv = append(argv, target_url)
cmd := exec.Command("curl", argv...)
err = cmd.Run()
if err != nil {
return
}
return
}