-
Notifications
You must be signed in to change notification settings - Fork 1
/
diskUse.go
298 lines (237 loc) · 8.54 KB
/
diskUse.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
// Copyright (c) 2016 OpenM++
// This code is licensed under the MIT license (see LICENSE.txt for details)
package main
import (
"io/fs"
"path/filepath"
"time"
"github.com/openmpp/go/ompp/config"
"github.com/openmpp/go/ompp/helper"
"github.com/openmpp/go/ompp/omppLog"
)
const diskScanDefaultInterval = 383 * 1000 // timeout in msec, default value of sleep interval between scanning storage use
const minDiskScanInterval = 11 * 1000 // timeout in msec, minimum value to sleep between scanning storage use
// storage space use state
type diskUseState struct {
IsOver bool // if true then storage use reach the limit
diskUseConfig // storage use settings
AllSize int64 // all oms instances size
TotalSize int64 // total size: models/bin size + download + upload
BinSize int64 // total models/bin size
DbSize int64 // total size of all db files
DownSize int64 // download total size
UpSize int64 // upload total size
UpdateTs int64 // info update time (unix milliseconds)
}
// model and database file info
type dbDiskUse struct {
Digest string // model digest
Size int64 // bytes, if non zero then model database file size
ModTs int64 // db file mod time info update time (unix milliseconds)
}
// storage usage control settings
type diskUseConfig struct {
DiskScanMs int64 // timeout in msec, sleep interval between scanning storage
Limit int64 // bytes, this instance storage limit
AllLimit int64 // bytes, total storage limit for all oms instances
}
/*
scan all models directories, download and upload directories to collect storage space usage.
Other file statistics also collected, e.g. files count, SQLite db file size for each model, etc.
*/
func scanDisk(doneC <-chan bool, refreshC <-chan bool) {
if !theCfg.isDiskUse {
return // storage use control disabled
}
// if disk use file does not updated more than 3 times of scan interval (and minimum 1 minute) then oms instance is dead
// disk use file: disk-#-_4040-#-size-#-100-#-status-#-ok-#-2022_07_08_23_45_12_123-#-125678.json
diskUsePtrn := filepath.Join(theCfg.jobDir, "state", "disk-#-"+theCfg.omsName+"-#-size-#-*-#-status-#-*-#-*-#-*.json")
// path to disk.ini: storage quotas and configuration
diskIniPath := filepath.Join(theCfg.jobDir, "disk.ini")
duState := diskUseState{
diskUseConfig: diskUseConfig{
DiskScanMs: diskScanDefaultInterval,
},
}
dbUse := []dbDiskUse{}
for {
isOk, cfg := initDiskState(diskIniPath)
if !isOk {
if isExitSleep(diskScanDefaultInterval, doneC) { // wait for doneC or sleep
return
}
continue
}
// clean previous state
nowTime := time.Now()
nowTs := nowTime.UnixMilli()
duState = diskUseState{
diskUseConfig: cfg,
UpdateTs: nowTs,
}
// find all disk use files and calculate total disk usage by all other oms instances
// if disk use file does not updated more than 3 times of scan interval (and minimum 1 minute) then oms instance is dead
minuteTs := nowTime.Add(-1 * time.Minute).UnixMilli()
minTs := nowTime.Add(-1 * 3 * time.Duration(cfg.DiskScanMs) * time.Millisecond).UnixMilli()
if minTs > minuteTs {
minTs = minuteTs
}
var nOtherSize int64 // all other oms instances disk use size
diskUseFiles := filesByPattern(diskUsePtrn, "Error at disk use files search")
for _, fp := range diskUseFiles {
oms, size, _, _, ts := parseDiskUseStatePath(fp)
if oms == "" || oms == theCfg.omsName {
continue // skip: invalid disk use state file path or it is current instance
}
if ts > minTs {
nOtherSize += size // oms instance is alive
}
}
// for all models get database file size and mod time
mbs := theCatalog.allModels()
if len(dbUse) != len(mbs) {
dbUse = make([]dbDiskUse, len(mbs)) // model catalog updated
}
for k := 0; k < len(mbs); k++ {
dbUse[k].Digest = mbs[k].model.Digest
dbUse[k].Size = 0
if st, e := fileStat(mbs[k].dbPath); e == nil { // skip on file errors
dbUse[k].Size = st.Size()
dbUse[k].ModTs = st.ModTime().UnixMilli()
}
duState.DbSize = duState.DbSize + dbUse[k].Size
}
// get total size of all files in the folder and sub-folders
doTotalSize := func(folderPath string) int64 {
var nTotal int64
err := filepath.Walk(folderPath, func(path string, fi fs.FileInfo, err error) error {
if err != nil {
omppLog.Log("Error at directory walk: ", path, " : ", err.Error())
return err
}
if !fi.IsDir() {
nTotal = nTotal + fi.Size()
}
return nil
})
if err != nil {
// omppLog.Log("Error at directory walk: ", folderPath, " :", err.Error())
}
return nTotal
}
// total size of models/bin directory, downlod and upload
mDir, _ := theCatalog.getModelDir()
duState.BinSize = doTotalSize(mDir)
if theCfg.downloadDir != "" {
duState.DownSize = doTotalSize(theCfg.downloadDir)
}
if theCfg.uploadDir != "" {
duState.UpSize = doTotalSize(theCfg.uploadDir)
}
duState.TotalSize = duState.BinSize + duState.DownSize + duState.UpSize
duState.AllSize = duState.TotalSize + nOtherSize
// check if current disk usage reach the limit
duState.IsOver = duState.Limit > 0 && duState.TotalSize >= cfg.Limit ||
cfg.AllLimit > 0 && duState.AllSize >= cfg.AllLimit
// update run catalog with current storage use state and save persistent part of the state
theRunCatalog.updateDiskUse(&duState, dbUse)
diskUseStateWrite(&duState, dbUse)
// wait for doneC or sleep
isExit := false
select {
case <-doneC:
isExit = true
case <-refreshC:
case <-time.After(time.Duration(cfg.DiskScanMs) * time.Millisecond):
}
if isExit {
break
}
}
}
// update run catalog with current disk use state
func (rsc *RunCatalog) updateDiskUse(duState *diskUseState, dbUse []dbDiskUse) {
rsc.rscLock.Lock()
defer rsc.rscLock.Unlock()
rsc.DiskUse = *duState
// copy db file info for current models list
if len(rsc.DbDiskUse) != len(rsc.models) {
rsc.DbDiskUse = make([]dbDiskUse, len(rsc.models))
}
k := 0
for dgst := range rsc.models {
rsc.DbDiskUse[k] = dbDiskUse{Digest: dgst}
for j := 0; j < len(dbUse); j++ {
if dbUse[j].Digest == dgst {
rsc.DbDiskUse[k] = dbUse[j]
break
}
}
k++
}
}
// return disk use status: flag is disk use over limit and disk use config
func (rsc *RunCatalog) getDiskUseStatus() (bool, diskUseConfig) {
rsc.rscLock.Lock()
defer rsc.rscLock.Unlock()
return rsc.DiskUse.IsOver, rsc.DiskUse.diskUseConfig
}
// return copy of current disk use state
func (rsc *RunCatalog) getDiskUse() (diskUseState, []dbDiskUse) {
rsc.rscLock.Lock()
defer rsc.rscLock.Unlock()
duState := rsc.DiskUse
dbUse := make([]dbDiskUse, len(rsc.DbDiskUse))
copy(dbUse, rsc.DbDiskUse)
return duState, dbUse
}
// read job service state and computational servers definition from job.ini
func initDiskState(diskIniPath string) (bool, diskUseConfig) {
cfg := diskUseConfig{DiskScanMs: diskScanDefaultInterval}
// read available resources limits and computational servers configuration from job.ini
if diskIniPath == "" || !fileExist(diskIniPath) {
return false, cfg
}
opts, err := config.FromIni(diskIniPath, theCfg.codePage)
if err != nil {
omppLog.Log(err)
return false, cfg
}
cfg.DiskScanMs = 1000 * opts.Int64("Common.ScanInterval", diskScanDefaultInterval)
if cfg.DiskScanMs < minDiskScanInterval {
cfg.DiskScanMs = diskScanDefaultInterval // if too low then use default
}
cfg.AllLimit = 1024 * 1024 * 1024 * opts.Int64("Common.AllUsersLimit", 0) // total limit in bytes for all oms instances
if cfg.AllLimit < 0 {
cfg.AllLimit = 0 // unlimited
}
// find oms instance limit defined by name
var uGb int64
isOk := opts.IsExist(theCfg.omsName + ".UserLimit")
if isOk {
uGb = opts.Int64(theCfg.omsName+".UserLimit", 0) // limit defined for current instance name
}
if !isOk && opts.IsExist("Common.Groups") {
gLst := helper.ParseCsvLine(opts.String("Common.Groups"), ',')
for k := 0; !isOk && k < len(gLst); k++ {
if !opts.IsExist(gLst[k]+".Users") || !opts.IsExist(gLst[k]+".UserLimit") {
continue // skip: no users in that group or no user limit defined
}
uLst := helper.ParseCsvLine(opts.String(gLst[k]+".Users"), ',')
for j := 0; !isOk && j < len(uLst); j++ {
isOk = uLst[j] == theCfg.omsName // check if instance name exists in that group
}
if isOk {
uGb = opts.Int64(gLst[k]+".UserLimit", 0) // group limit applied to the current instance
}
}
}
if !isOk {
uGb = opts.Int64("Common.UserLimit", 0) // apply limit common for any user
}
if uGb < 0 {
uGb = 0
}
cfg.Limit = 1024 * 1024 * 1024 * uGb // bytes, storage limit for current instance name
return true, cfg
}