forked from influxdata/influxdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtx.go
270 lines (223 loc) · 5.89 KB
/
tx.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
package influxdb
import (
"bytes"
"fmt"
"sync"
"time"
"github.com/boltdb/bolt"
"github.com/influxdb/influxdb/influxql"
)
// tx represents a transaction that spans multiple shard data stores.
// This transaction will open and close all data stores atomically.
type tx struct {
mu sync.Mutex
server *Server
opened bool
now time.Time
itrs []*shardIterator // shard iterators
}
// newTx return a new initialized Tx.
func newTx(server *Server) *tx {
return &tx{
server: server,
now: time.Now(),
}
}
// SetNow sets the current time for the transaction.
func (tx *tx) SetNow(now time.Time) { tx.now = now }
// Open opens a read-only transaction on all data stores atomically.
func (tx *tx) Open() error {
tx.mu.Lock()
defer tx.mu.Unlock()
// Mark transaction as open.
tx.opened = true
// Open each iterator individually. If any fail close the transaction and error out
for _, itr := range tx.itrs {
if err := itr.open(); err != nil {
_ = tx.close()
return err
}
}
return nil
}
// Close closes all data store transactions atomically.
func (tx *tx) Close() error {
tx.mu.Lock()
defer tx.mu.Unlock()
return tx.close()
}
func (tx *tx) close() error {
// Mark transaction as closed.
tx.opened = false
for _, itr := range tx.itrs {
_ = itr.close()
}
return nil
}
// CreateIterators returns an iterator for a simple select statement.
func (tx *tx) CreateIterators(stmt *influxql.SelectStatement) ([]influxql.Iterator, error) {
// Parse the source segments.
database, policyName, measurement, err := splitIdent(stmt.Source.(*influxql.Measurement).Name)
if err != nil {
return nil, err
}
// Grab time range from statement.
tmin, tmax := influxql.TimeRange(stmt.Condition)
if tmin.IsZero() {
tmin = time.Unix(0, 1)
}
if tmax.IsZero() {
tmax = tx.now
}
// Find database and retention policy.
db := tx.server.databases[database]
if db == nil {
return nil, ErrDatabaseNotFound
}
rp := db.policies[policyName]
if rp == nil {
return nil, ErrRetentionPolicyNotFound
}
// Find shard groups within time range.
var shardGroups []*ShardGroup
for _, group := range rp.shardGroups {
if timeBetweenInclusive(group.StartTime, tmin, tmax) || timeBetweenInclusive(group.EndTime, tmin, tmax) {
shardGroups = append(shardGroups, group)
}
}
if len(shardGroups) == 0 {
return nil, nil
}
// Normalize dimensions to extract the interval.
_, dimensions, err := stmt.Dimensions.Normalize()
if err != nil {
return nil, err
}
// Find measurement.
m, err := tx.server.measurement(database, measurement)
if err != nil {
return nil, err
}
// Find field.
fieldName := stmt.Fields[0].Expr.(*influxql.VarRef).Val
f := m.FieldByName(fieldName)
if f == nil {
return nil, fmt.Errorf("field not found: %s", fieldName)
}
tagSets := m.tagSets(stmt, dimensions)
// Convert time range to bytes.
kmin := u64tob(uint64(tmin.UnixNano()))
kmax := u64tob(uint64(tmax.UnixNano()))
// Create an iterator for every shard.
var itrs []influxql.Iterator
for tag, set := range tagSets {
for _, group := range shardGroups {
// TODO: only create iterators for the shards we actually have to hit in a group
for _, sh := range group.Shards {
itr := &shardIterator{
fieldID: f.ID,
tags: tag,
conditions: set, // TODO: only pass in conditions for series that are in this shard
db: sh.store,
cur: &multiCursor{
kmin: kmin,
kmax: kmax,
},
}
// Add to tx so the bolt transaction can be opened/closed.
tx.itrs = append(tx.itrs, itr)
itrs = append(itrs, itr)
}
}
}
return itrs, nil
}
// splitIdent splits an identifier into it's database, policy, and measurement parts.
func splitIdent(s string) (db, rp, m string, err error) {
a, err := influxql.SplitIdent(s)
if err != nil {
return "", "", "", err
} else if len(a) != 3 {
return "", "", "", fmt.Errorf("invalid ident, expected 3 segments: %q", s)
}
return a[0], a[1], a[2], nil
}
// shardIterator represents an iterator for traversing over a single series.
type shardIterator struct {
fieldID uint8
tags string // encoded dimensional tag values
conditions map[uint32]influxql.Expr
db *bolt.DB // data stores by shard id
txn *bolt.Tx // read transactions by shard id
cur *multiCursor
}
func (i *shardIterator) Tags() string { return i.tags }
func (i *shardIterator) Next() (key int64, value interface{}) {
// Ignore if there's no more data.
k, v := i.cur.Next()
if k == nil {
return 0, nil
}
// Read timestamp & field value.
key = int64(btou64(k))
value = unmarshalValue(v, i.fieldID)
return key, value
}
func (i *shardIterator) open() error {
// Open the data store
txn, err := i.db.Begin(false)
if err != nil {
return err
}
i.txn = txn
// Open cursors for each series id
for id, _ := range i.conditions {
b := i.txn.Bucket(u32tob(id))
if b == nil {
continue
}
// add the cursor fo this series on the shard
i.cur.cursors = append(i.cur.cursors, b.Cursor())
}
i.cur.initialize()
return nil
}
func (i *shardIterator) close() error {
_ = i.txn.Rollback()
return nil
}
type multiCursor struct {
cursors []*bolt.Cursor
keyValues []keyValue
kmin, kmax []byte // min/max keys
}
type keyValue struct {
key []byte
value []byte
}
func (c *multiCursor) initialize() {
c.keyValues = make([]keyValue, len(c.cursors))
for i, cur := range c.cursors {
c.keyValues[i].key, c.keyValues[i].value = cur.Seek(c.kmin)
}
}
func (c *multiCursor) Next() (key, value []byte) {
min := -1
for i, kv := range c.keyValues {
if kv.key != nil && bytes.Compare(kv.key, c.kmax) == -1 {
min = i
}
}
// if min is -1 we've exhausted all cursors for the given time range
if min == -1 {
return nil, nil
}
kv := c.keyValues[min]
key = kv.key
value = kv.value
c.keyValues[min].key, c.keyValues[min].value = c.cursors[min].Next()
if bytes.Compare(c.keyValues[min].key, c.kmax) == 1 {
c.keyValues[min].key = nil
}
return
}