drainer/*: get latest timestamp from pd when initial-commit-ts is -1 (#787)

drainer/config.go

@@ -123,7 +123,7 @@ func NewConfig() *Config {
  fs.StringVar(&cfg.MetricsAddr, "metrics-addr", "", "prometheus pushgateway address, leaves it empty will disable prometheus push")
  fs.IntVar(&cfg.MetricsInterval, "metrics-interval", 15, "prometheus client push interval in second, set \"0\" to disable prometheus push")
  fs.StringVar(&cfg.LogFile, "log-file", "", "log file path")
- fs.Int64Var(&cfg.InitialCommitTS, "initial-commit-ts", 0, "if drainer donesn't have checkpoint, use initial commitTS to initial checkpoint")
+ fs.Int64Var(&cfg.InitialCommitTS, "initial-commit-ts", -1, "if drainer donesn't have checkpoint, use initial commitTS to initial checkpoint, will get a latest timestamp from pd if setting to be -1")
  fs.StringVar(&cfg.Compressor, "compressor", "", "use the specified compressor to compress payload between pump and drainer, only 'gzip' is supported now (default \"\", ie. compression disabled.)")
  fs.IntVar(&cfg.SyncerCfg.TxnBatch, "txn-batch", 20, "number of binlog events in a transaction batch")
  fs.StringVar(&cfg.SyncerCfg.IgnoreSchemas, "ignore-schemas", "INFORMATION_SCHEMA,PERFORMANCE_SCHEMA,mysql", "disable sync those schemas")

drainer/server.go

@@ -116,6 +116,11 @@ func NewServer(cfg *Config) (*Server, error) {
  }
  latestTime := time.Now()
 
+ if cfg.InitialCommitTS == -1 {
+ log.Info("set InitialCommitTS", zap.Int64("ts", latestTS))
+ cfg.InitialCommitTS = latestTS
+ }
+
  cfg.SyncerCfg.To.ClusterID = clusterID
  pdCli.Close()
 

tests/dailytest/case.go

@@ -211,7 +211,8 @@ func RunCase(src *sql.DB, dst *sql.DB, schema string) {
  // run casePKAddDuplicateUK
  tr.run(func(src *sql.DB) {
  err := execSQLs(src, casePKAddDuplicateUK)
- if err != nil && !strings.Contains(err.Error(), "Duplicate for key") {
+ // the add unique index will failed by duplicate entry
+ if err != nil && !strings.Contains(err.Error(), "Duplicate") {
  log.S().Fatal(err)
  }
  })

tests/kafka/run.sh

@@ -2,13 +2,9 @@
 
 set -e
 
-# use latest ts as initial-commit-ts, so we can skip binlog by previous test case
-ms=$(date +'%s')
-ts=$(($ms*1000<<18))
-
 cd "$(dirname "$0")"
 
-args="-initial-commit-ts=$ts"
+args="-initial-commit-ts=-1"
 
 kafka_addr=${KAFKA_ADDRS-127.0.0.1:9092}
 

tests/reparo/run.sh

@@ -5,18 +5,17 @@ set -e
 cd "$(dirname "$0")"
 
 # use latest ts as initial-commit-ts, so we can skip binlog by previous test case
-ms=$(date +'%s')
-ts=$(($ms*1000<<18))
-args="-initial-commit-ts=$ts"
+args="-initial-commit-ts=-1"
 down_run_sql "DROP DATABASE IF EXISTS tidb_binlog"
-run_sql "CREATE DATABASE IF NOT EXISTS \`reparo_test\`"
 
 rm -rf /tmp/tidb_binlog_test/data.drainer
 
 run_drainer "$args" &
 
 GO111MODULE=on go build -o out
 
+run_sql "CREATE DATABASE IF NOT EXISTS \`reparo_test\`"
+
 ./out -config ./config.toml > ${OUT_DIR-/tmp}/$TEST_NAME.out 2>&1
 
 sleep 5

tests/restart/run.sh

@@ -9,9 +9,7 @@ STATUS_LOG="${OUT_DIR}/status.log"
 
 # run drainer, and drainer's status should be online
 # use latest ts as initial-commit-ts, so we can skip binlog by previous test case
-ms=$(date +'%s')
-ts=$(($ms*1000<<18))
-args="-initial-commit-ts=$ts"
+args="-initial-commit-ts=-1"
 down_run_sql "DROP DATABASE IF EXISTS tidb_binlog"
 rm -rf /tmp/tidb_binlog_test/data.drainer
 

tests/status/run.sh

@@ -9,9 +9,7 @@ OUT_DIR=/tmp/tidb_binlog_test
 STATUS_LOG="${OUT_DIR}/status.log"
 
 # use latest ts as initial-commit-ts, so we can skip binlog by previous test case
-ms=$(date +'%s')
-ts=$(($ms*1000<<18))
-args="-initial-commit-ts=$ts"
+args="-initial-commit-ts=-1"
 down_run_sql "DROP DATABASE IF EXISTS tidb_binlog"
 rm -rf /tmp/tidb_binlog_test/data.drainer