Beispiel #1
0
func (d *ddl) handleJobQueue() error {
	for {
		if d.isClosed() {
			return nil
		}

		waitTime := 2 * d.lease

		var job *model.Job
		err := kv.RunInNewTxn(d.store, false, func(txn kv.Transaction) error {
			t := meta.NewMeta(txn)
			owner, err := d.checkOwner(t)
			if terror.ErrorEqual(err, ErrNotOwner) {
				// we are not owner, return and retry checking later.
				return nil
			} else if err != nil {
				return errors.Trace(err)
			}

			// become the owner
			// get the first job and run
			job, err = d.getFirstJob(t)
			if job == nil || err != nil {
				return errors.Trace(err)
			}

			if job.IsRunning() {
				// if we enter a new state, crash when waiting 2 * lease time, and restart quickly,
				// we may run the job immediately again, but we don't wait enough 2 * lease time to
				// let other servers update the schema.
				// so here we must check the elapsed time from last update, if < 2 * lease, we must
				// wait again.
				elapsed := time.Duration(time.Now().UnixNano() - job.LastUpdateTS)
				if elapsed > 0 && elapsed < waitTime {
					log.Warnf("the elapsed time from last update is %s < %s, wait again", elapsed, waitTime)
					waitTime -= elapsed
					return nil
				}
			}

			log.Warnf("run DDL job %v", job)

			d.hook.OnJobRunBefore(job)

			// if run job meets error, we will save this error in job Error
			// and retry later if the job is not cancelled.
			d.runJob(t, job)

			if job.IsFinished() {
				err = d.finishJob(t, job)
			} else {
				err = d.updateJob(t, job)
			}

			if err != nil {
				return errors.Trace(err)
			}

			// running job may cost some time, so here we must update owner status to
			// prevent other become the owner.
			owner.LastUpdateTS = time.Now().UnixNano()
			if err = t.SetDDLOwner(owner); err != nil {
				return errors.Trace(err)
			}
			return errors.Trace(err)
		})

		if err != nil {
			return errors.Trace(err)
		} else if job == nil {
			// no job now, return and retry get later.
			return nil
		}

		d.hook.OnJobUpdated(job)

		// here means the job enters another state (delete only, write only, public, etc...) or is cancelled.
		// if the job is done or still running, we will wait 2 * lease time to guarantee other servers to update
		// the newest schema.
		if job.State == model.JobRunning || job.State == model.JobDone {
			d.waitSchemaChanged(waitTime)
		}

		if job.IsFinished() {
			asyncNotify(d.jobDoneCh)
		}
	}
}