main.go 6.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205
  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. "flag"
  17. "fmt"
  18. "net/http"
  19. "os"
  20. "strings"
  21. "github.com/coreos/pkg/capnslog"
  22. "github.com/prometheus/client_golang/prometheus"
  23. )
  24. var plog = capnslog.NewPackageLogger("github.com/coreos/etcd", "etcd-tester")
  25. const (
  26. defaultClientPort = 2379
  27. defaultPeerPort = 2380
  28. defaultFailpointPort = 2381
  29. )
  30. func main() {
  31. endpointStr := flag.String("agent-endpoints", "localhost:9027", "HTTP RPC endpoints of agents. Do not specify the schema.")
  32. clientPorts := flag.String("client-ports", "", "etcd client port for each agent endpoint")
  33. peerPorts := flag.String("peer-ports", "", "etcd peer port for each agent endpoint")
  34. failpointPorts := flag.String("failpoint-ports", "", "etcd failpoint port for each agent endpoint")
  35. datadir := flag.String("data-dir", "agent.etcd", "etcd data directory location on agent machine.")
  36. stressKeyLargeSize := flag.Uint("stress-key-large-size", 32*1024+1, "the size of each large key written into etcd.")
  37. stressKeySize := flag.Uint("stress-key-size", 100, "the size of each small key written into etcd.")
  38. stressKeySuffixRange := flag.Uint("stress-key-count", 250000, "the count of key range written into etcd.")
  39. limit := flag.Int("limit", -1, "the limit of rounds to run failure set (-1 to run without limits).")
  40. stressQPS := flag.Int("stress-qps", 10000, "maximum number of stresser requests per second.")
  41. schedCases := flag.String("schedule-cases", "", "test case schedule")
  42. consistencyCheck := flag.Bool("consistency-check", true, "true to check consistency (revision, hash)")
  43. isV2Only := flag.Bool("v2-only", false, "'true' to run V2 only tester.")
  44. stresserType := flag.String("stresser", "default", "specify stresser (\"default\" or \"nop\").")
  45. failureTypes := flag.String("failures", "default,failpoints", "specify failures (concat of \"default\" and \"failpoints\").")
  46. externalFailures := flag.String("external-failures", "", "specify a path of script for enabling/disabling an external fault injector")
  47. flag.Parse()
  48. eps := strings.Split(*endpointStr, ",")
  49. cports := portsFromArg(*clientPorts, len(eps), defaultClientPort)
  50. pports := portsFromArg(*peerPorts, len(eps), defaultPeerPort)
  51. fports := portsFromArg(*failpointPorts, len(eps), defaultFailpointPort)
  52. agents := make([]agentConfig, len(eps))
  53. for i := range eps {
  54. agents[i].endpoint = eps[i]
  55. agents[i].clientPort = cports[i]
  56. agents[i].peerPort = pports[i]
  57. agents[i].failpointPort = fports[i]
  58. agents[i].datadir = *datadir
  59. }
  60. sConfig := &stressConfig{
  61. qps: *stressQPS,
  62. keyLargeSize: int(*stressKeyLargeSize),
  63. keySize: int(*stressKeySize),
  64. keySuffixRange: int(*stressKeySuffixRange),
  65. v2: *isV2Only,
  66. }
  67. lsConfig := &leaseStressConfig{
  68. numLeases: 10,
  69. keysPerLease: 10,
  70. qps: *stressQPS, // only used to create nop stresser in leaseStresserBuilder
  71. }
  72. c := &cluster{
  73. agents: agents,
  74. v2Only: *isV2Only,
  75. stressBuilder: newStressBuilder(*stresserType, sConfig),
  76. leaseStresserBuilder: newLeaseStresserBuilder(*stresserType, lsConfig),
  77. consistencyCheck: *consistencyCheck,
  78. }
  79. if err := c.bootstrap(); err != nil {
  80. plog.Fatal(err)
  81. }
  82. defer c.Terminate()
  83. // ensure cluster is fully booted to know failpoints are available
  84. c.WaitHealth()
  85. var failures []failure
  86. if failureTypes != nil && *failureTypes != "" {
  87. failures = makeFailures(*failureTypes, c)
  88. }
  89. if externalFailures != nil && *externalFailures != "" {
  90. if len(failures) != 0 {
  91. plog.Errorf("specify only one of -failures or -external-failures")
  92. os.Exit(1)
  93. }
  94. failures = append(failures, newFailureExternal(*externalFailures))
  95. }
  96. if len(failures) == 0 {
  97. plog.Infof("no failures\n")
  98. failures = append(failures, newFailureNop())
  99. }
  100. schedule := failures
  101. if schedCases != nil && *schedCases != "" {
  102. cases := strings.Split(*schedCases, " ")
  103. schedule = make([]failure, len(cases))
  104. for i := range cases {
  105. caseNum := 0
  106. n, err := fmt.Sscanf(cases[i], "%d", &caseNum)
  107. if n == 0 || err != nil {
  108. plog.Fatalf(`couldn't parse case "%s" (%v)`, cases[i], err)
  109. }
  110. schedule[i] = failures[caseNum]
  111. }
  112. }
  113. t := &tester{
  114. failures: schedule,
  115. cluster: c,
  116. limit: *limit,
  117. }
  118. sh := statusHandler{status: &t.status}
  119. http.Handle("/status", sh)
  120. http.Handle("/metrics", prometheus.Handler())
  121. go func() { plog.Fatal(http.ListenAndServe(":9028", nil)) }()
  122. t.runLoop()
  123. }
  124. // portsFromArg converts a comma separated list into a slice of ints
  125. func portsFromArg(arg string, n, defaultPort int) []int {
  126. ret := make([]int, n)
  127. if len(arg) == 0 {
  128. for i := range ret {
  129. ret[i] = defaultPort
  130. }
  131. return ret
  132. }
  133. s := strings.Split(arg, ",")
  134. if len(s) != n {
  135. fmt.Printf("expected %d ports, got %d (%s)\n", n, len(s), arg)
  136. os.Exit(1)
  137. }
  138. for i := range s {
  139. if _, err := fmt.Sscanf(s[i], "%d", &ret[i]); err != nil {
  140. fmt.Println(err)
  141. os.Exit(1)
  142. }
  143. }
  144. return ret
  145. }
  146. func makeFailures(types string, c *cluster) []failure {
  147. var failures []failure
  148. fails := strings.Split(types, ",")
  149. for i := range fails {
  150. switch fails[i] {
  151. case "default":
  152. defaultFailures := []failure{
  153. newFailureKillAll(),
  154. newFailureKillMajority(),
  155. newFailureKillOne(),
  156. newFailureKillLeader(),
  157. newFailureKillOneForLongTime(),
  158. newFailureKillLeaderForLongTime(),
  159. newFailureIsolate(),
  160. newFailureIsolateAll(),
  161. newFailureSlowNetworkOneMember(),
  162. newFailureSlowNetworkLeader(),
  163. newFailureSlowNetworkAll(),
  164. }
  165. failures = append(failures, defaultFailures...)
  166. case "failpoints":
  167. fpFailures, fperr := failpointFailures(c)
  168. if len(fpFailures) == 0 {
  169. plog.Infof("no failpoints found (%v)", fperr)
  170. }
  171. failures = append(failures, fpFailures...)
  172. default:
  173. plog.Errorf("unknown failure: %s\n", fails[i])
  174. os.Exit(1)
  175. }
  176. }
  177. return failures
  178. }