tester.go 6.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242
  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 ok, err := tt.doRound(round); !ok {
  41. if err != nil {
  42. if tt.cleanup() != nil {
  43. return
  44. }
  45. }
  46. prevCompactRev = 0 // reset after clean up
  47. continue
  48. }
  49. // -1 so that logPrefix doesn't print out 'case'
  50. tt.status.setCase(-1)
  51. revToCompact := max(0, tt.currentRevision-10000)
  52. compactN := revToCompact - prevCompactRev
  53. timeout := 10 * time.Second
  54. if compactN > 0 {
  55. timeout += time.Duration(compactN/compactQPS) * time.Second
  56. }
  57. prevCompactRev = revToCompact
  58. plog.Printf("%s compacting %d entries (timeout %v)", tt.logPrefix(), compactN, timeout)
  59. if err := tt.compact(revToCompact, timeout); err != nil {
  60. plog.Warningf("%s functional-tester compact got error (%v)", tt.logPrefix(), err)
  61. if tt.cleanup() != nil {
  62. return
  63. }
  64. prevCompactRev = 0 // reset after clean up
  65. }
  66. if round > 0 && round%500 == 0 { // every 500 rounds
  67. if err := tt.defrag(); err != nil {
  68. plog.Warningf("%s functional-tester returning with error (%v)", tt.logPrefix(), err)
  69. return
  70. }
  71. }
  72. }
  73. plog.Printf("%s functional-tester is finished", tt.logPrefix())
  74. }
  75. func (tt *tester) doRound(round int) (bool, error) {
  76. for j, f := range tt.failures {
  77. caseTotalCounter.WithLabelValues(f.Desc()).Inc()
  78. tt.status.setCase(j)
  79. if err := tt.cluster.WaitHealth(); err != nil {
  80. plog.Printf("%s wait full health error: %v", tt.logPrefix(), err)
  81. return false, nil
  82. }
  83. plog.Printf("%s injecting failure %q", tt.logPrefix(), f.Desc())
  84. if err := f.Inject(tt.cluster, round); err != nil {
  85. plog.Printf("%s injection error: %v", tt.logPrefix(), err)
  86. return false, nil
  87. }
  88. plog.Printf("%s injected failure", tt.logPrefix())
  89. plog.Printf("%s recovering failure %q", tt.logPrefix(), f.Desc())
  90. if err := f.Recover(tt.cluster, round); err != nil {
  91. plog.Printf("%s recovery error: %v", tt.logPrefix(), err)
  92. return false, nil
  93. }
  94. plog.Printf("%s recovered failure", tt.logPrefix())
  95. if err := tt.checkConsistency(); err != nil {
  96. plog.Warningf("%s functional-tester returning with tt.checkConsistency error (%v)", tt.logPrefix(), err)
  97. return false, err
  98. }
  99. plog.Printf("%s succeed!", tt.logPrefix())
  100. }
  101. return true, nil
  102. }
  103. func (tt *tester) updateRevision() error {
  104. if tt.cluster.v2Only {
  105. return nil
  106. }
  107. revs, _, err := tt.cluster.getRevisionHash()
  108. for _, rev := range revs {
  109. tt.currentRevision = rev
  110. break // just need get one of the current revisions
  111. }
  112. plog.Printf("%s updated current revision to %d", tt.logPrefix(), tt.currentRevision)
  113. return err
  114. }
  115. func (tt *tester) checkConsistency() (err error) {
  116. tt.cancelStressers()
  117. defer func() {
  118. if err != nil {
  119. return
  120. }
  121. if err = tt.updateRevision(); err != nil {
  122. plog.Warningf("%s functional-tester returning with tt.updateRevision error (%v)", tt.logPrefix(), err)
  123. return
  124. }
  125. err = tt.startStressers()
  126. }()
  127. if err = tt.cluster.Checker.Check(); err != nil {
  128. plog.Printf("%s %v", tt.logPrefix(), err)
  129. }
  130. return err
  131. }
  132. func (tt *tester) compact(rev int64, timeout time.Duration) (err error) {
  133. tt.cancelStressers()
  134. defer func() {
  135. if err == nil {
  136. err = tt.startStressers()
  137. }
  138. }()
  139. plog.Printf("%s compacting storage (current revision %d, compact revision %d)", tt.logPrefix(), tt.currentRevision, rev)
  140. if err = tt.cluster.compactKV(rev, timeout); err != nil {
  141. return err
  142. }
  143. plog.Printf("%s compacted storage (compact revision %d)", tt.logPrefix(), rev)
  144. plog.Printf("%s checking compaction (compact revision %d)", tt.logPrefix(), rev)
  145. if err = tt.cluster.checkCompact(rev); err != nil {
  146. plog.Warningf("%s checkCompact error (%v)", tt.logPrefix(), err)
  147. return err
  148. }
  149. plog.Printf("%s confirmed compaction (compact revision %d)", tt.logPrefix(), rev)
  150. return nil
  151. }
  152. func (tt *tester) defrag() error {
  153. plog.Printf("%s defragmenting...", tt.logPrefix())
  154. if err := tt.cluster.defrag(); err != nil {
  155. plog.Warningf("%s defrag error (%v)", tt.logPrefix(), err)
  156. if cerr := tt.cleanup(); cerr != nil {
  157. return fmt.Errorf("%s, %s", err, cerr)
  158. }
  159. return err
  160. }
  161. plog.Printf("%s defragmented...", tt.logPrefix())
  162. return nil
  163. }
  164. func (tt *tester) logPrefix() string {
  165. var (
  166. rd = tt.status.getRound()
  167. cs = tt.status.getCase()
  168. prefix = fmt.Sprintf("[round#%d case#%d]", rd, cs)
  169. )
  170. if cs == -1 {
  171. prefix = fmt.Sprintf("[round#%d]", rd)
  172. }
  173. return prefix
  174. }
  175. func (tt *tester) cleanup() error {
  176. roundFailedTotalCounter.Inc()
  177. desc := "compact/defrag"
  178. if tt.status.Case != -1 {
  179. desc = tt.failures[tt.status.Case].Desc()
  180. }
  181. caseFailedTotalCounter.WithLabelValues(desc).Inc()
  182. plog.Printf("%s cleaning up...", tt.logPrefix())
  183. if err := tt.cluster.Cleanup(); err != nil {
  184. plog.Warningf("%s cleanup error: %v", tt.logPrefix(), err)
  185. return err
  186. }
  187. if err := tt.cluster.Reset(); err != nil {
  188. plog.Warningf("%s cleanup Bootstrap error: %v", tt.logPrefix(), err)
  189. return err
  190. }
  191. return nil
  192. }
  193. func (tt *tester) cancelStressers() {
  194. plog.Printf("%s canceling the stressers...", tt.logPrefix())
  195. for _, s := range tt.cluster.Stressers {
  196. s.Cancel()
  197. }
  198. plog.Printf("%s canceled stressers", tt.logPrefix())
  199. }
  200. func (tt *tester) startStressers() error {
  201. plog.Printf("%s starting the stressers...", tt.logPrefix())
  202. for _, s := range tt.cluster.Stressers {
  203. if err := s.Stress(); err != nil {
  204. return err
  205. }
  206. }
  207. plog.Printf("%s started stressers", tt.logPrefix())
  208. return nil
  209. }