tester.go 6.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241
  1. // Copyright 2015 The etcd Authors
  2. //
  3. // Licensed under the Apache License, Version 2.0 (the "License");
  4. // you may not use this file except in compliance with the License.
  5. // You may obtain a copy of the License at
  6. //
  7. // http://www.apache.org/licenses/LICENSE-2.0
  8. //
  9. // Unless required by applicable law or agreed to in writing, software
  10. // distributed under the License is distributed on an "AS IS" BASIS,
  11. // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12. // See the License for the specific language governing permissions and
  13. // limitations under the License.
  14. package main
  15. import (
  16. "fmt"
  17. "time"
  18. )
  19. type tester struct {
  20. failures []failure
  21. cluster *cluster
  22. limit int
  23. status Status
  24. currentRevision int64
  25. }
  26. // compactQPS is rough number of compact requests per second.
  27. // Previous tests showed etcd can compact about 60,000 entries per second.
  28. const compactQPS = 50000
  29. func (tt *tester) runLoop() {
  30. tt.status.Since = time.Now()
  31. tt.status.RoundLimit = tt.limit
  32. tt.status.cluster = tt.cluster
  33. for _, f := range tt.failures {
  34. tt.status.Failures = append(tt.status.Failures, f.Desc())
  35. }
  36. var prevCompactRev int64
  37. for round := 0; round < tt.limit || tt.limit == -1; round++ {
  38. tt.status.setRound(round)
  39. roundTotalCounter.Inc()
  40. if err := tt.doRound(round); err != nil {
  41. plog.Warningf("%s functional-tester returning with error (%v)", tt.logPrefix(), err)
  42. if tt.cleanup() != nil {
  43. return
  44. }
  45. prevCompactRev = 0 // reset after clean up
  46. continue
  47. }
  48. // -1 so that logPrefix doesn't print out 'case'
  49. tt.status.setCase(-1)
  50. revToCompact := max(0, tt.currentRevision-10000)
  51. compactN := revToCompact - prevCompactRev
  52. timeout := 10 * time.Second
  53. if compactN > 0 {
  54. timeout += time.Duration(compactN/compactQPS) * time.Second
  55. }
  56. prevCompactRev = revToCompact
  57. plog.Printf("%s compacting %d entries (timeout %v)", tt.logPrefix(), compactN, timeout)
  58. if err := tt.compact(revToCompact, timeout); err != nil {
  59. plog.Warningf("%s functional-tester compact got error (%v)", tt.logPrefix(), err)
  60. if tt.cleanup() != nil {
  61. return
  62. }
  63. prevCompactRev = 0 // reset after clean up
  64. }
  65. if round > 0 && round%500 == 0 { // every 500 rounds
  66. if err := tt.defrag(); err != nil {
  67. plog.Warningf("%s functional-tester returning with error (%v)", tt.logPrefix(), err)
  68. return
  69. }
  70. }
  71. }
  72. plog.Printf("%s functional-tester is finished", tt.logPrefix())
  73. }
  74. func (tt *tester) doRound(round int) error {
  75. for j, f := range tt.failures {
  76. caseTotalCounter.WithLabelValues(f.Desc()).Inc()
  77. tt.status.setCase(j)
  78. if err := tt.cluster.WaitHealth(); err != nil {
  79. return fmt.Errorf("wait full health error: %v", err)
  80. }
  81. plog.Printf("%s injecting failure %q", tt.logPrefix(), f.Desc())
  82. if err := f.Inject(tt.cluster, round); err != nil {
  83. return fmt.Errorf("injection error: %v", err)
  84. }
  85. plog.Printf("%s injected failure", tt.logPrefix())
  86. plog.Printf("%s recovering failure %q", tt.logPrefix(), f.Desc())
  87. if err := f.Recover(tt.cluster, round); err != nil {
  88. return fmt.Errorf("recovery error: %v", err)
  89. }
  90. plog.Printf("%s wait until cluster is healthy", tt.logPrefix())
  91. if err := tt.cluster.WaitHealth(); err != nil {
  92. return fmt.Errorf("wait full health error: %v", err)
  93. }
  94. plog.Printf("%s recovered failure", tt.logPrefix())
  95. if err := tt.checkConsistency(); err != nil {
  96. return fmt.Errorf("tt.checkConsistency error (%v)", err)
  97. }
  98. plog.Printf("%s succeed!", tt.logPrefix())
  99. }
  100. return nil
  101. }
  102. func (tt *tester) updateRevision() error {
  103. if tt.cluster.v2Only {
  104. return nil
  105. }
  106. revs, _, err := tt.cluster.getRevisionHash()
  107. for _, rev := range revs {
  108. tt.currentRevision = rev
  109. break // just need get one of the current revisions
  110. }
  111. plog.Printf("%s updated current revision to %d", tt.logPrefix(), tt.currentRevision)
  112. return err
  113. }
  114. func (tt *tester) checkConsistency() (err error) {
  115. tt.cancelStressers()
  116. defer func() {
  117. if err != nil {
  118. return
  119. }
  120. if err = tt.updateRevision(); err != nil {
  121. plog.Warningf("%s functional-tester returning with tt.updateRevision error (%v)", tt.logPrefix(), err)
  122. return
  123. }
  124. err = tt.startStressers()
  125. }()
  126. if err = tt.cluster.Checker.Check(); err != nil {
  127. plog.Printf("%s %v", tt.logPrefix(), err)
  128. }
  129. return err
  130. }
  131. func (tt *tester) compact(rev int64, timeout time.Duration) (err error) {
  132. tt.cancelStressers()
  133. defer func() {
  134. if err == nil {
  135. err = tt.startStressers()
  136. }
  137. }()
  138. plog.Printf("%s compacting storage (current revision %d, compact revision %d)", tt.logPrefix(), tt.currentRevision, rev)
  139. if err = tt.cluster.compactKV(rev, timeout); err != nil {
  140. return err
  141. }
  142. plog.Printf("%s compacted storage (compact revision %d)", tt.logPrefix(), rev)
  143. plog.Printf("%s checking compaction (compact revision %d)", tt.logPrefix(), rev)
  144. if err = tt.cluster.checkCompact(rev); err != nil {
  145. plog.Warningf("%s checkCompact error (%v)", tt.logPrefix(), err)
  146. return err
  147. }
  148. plog.Printf("%s confirmed compaction (compact revision %d)", tt.logPrefix(), rev)
  149. return nil
  150. }
  151. func (tt *tester) defrag() error {
  152. plog.Printf("%s defragmenting...", tt.logPrefix())
  153. if err := tt.cluster.defrag(); err != nil {
  154. plog.Warningf("%s defrag error (%v)", tt.logPrefix(), err)
  155. if cerr := tt.cleanup(); cerr != nil {
  156. return fmt.Errorf("%s, %s", err, cerr)
  157. }
  158. return err
  159. }
  160. plog.Printf("%s defragmented...", tt.logPrefix())
  161. return nil
  162. }
  163. func (tt *tester) logPrefix() string {
  164. var (
  165. rd = tt.status.getRound()
  166. cs = tt.status.getCase()
  167. prefix = fmt.Sprintf("[round#%d case#%d]", rd, cs)
  168. )
  169. if cs == -1 {
  170. prefix = fmt.Sprintf("[round#%d]", rd)
  171. }
  172. return prefix
  173. }
  174. func (tt *tester) cleanup() error {
  175. roundFailedTotalCounter.Inc()
  176. desc := "compact/defrag"
  177. if tt.status.Case != -1 {
  178. desc = tt.failures[tt.status.Case].Desc()
  179. }
  180. caseFailedTotalCounter.WithLabelValues(desc).Inc()
  181. plog.Printf("%s cleaning up...", tt.logPrefix())
  182. if err := tt.cluster.Cleanup(); err != nil {
  183. plog.Warningf("%s cleanup error: %v", tt.logPrefix(), err)
  184. return err
  185. }
  186. if err := tt.cluster.Reset(); err != nil {
  187. plog.Warningf("%s cleanup Bootstrap error: %v", tt.logPrefix(), err)
  188. return err
  189. }
  190. return nil
  191. }
  192. func (tt *tester) cancelStressers() {
  193. plog.Printf("%s canceling the stressers...", tt.logPrefix())
  194. for _, s := range tt.cluster.Stressers {
  195. s.Cancel()
  196. }
  197. plog.Printf("%s canceled stressers", tt.logPrefix())
  198. }
  199. func (tt *tester) startStressers() error {
  200. plog.Printf("%s starting the stressers...", tt.logPrefix())
  201. for _, s := range tt.cluster.Stressers {
  202. if err := s.Stress(); err != nil {
  203. return err
  204. }
  205. }
  206. plog.Printf("%s started stressers", tt.logPrefix())
  207. return nil
  208. }