Parallel building cop tasks with cop workers in TiKV client. (select * from limit 1 is too slow)

[SunRunAway]

Description

The current procedure when sending a cop request is

1. build all cop tasks in memory.
2. send each cop task to copIteratorWorker
3. copIteratorWorker execute the task

If we have thousands of tasks, the building step may take too long(and it may load region info from pd), it will enlarge the time of first meeting row.

We need build all cop tasks in a diffrent goroutine, and send tasks which successfully builded immediately to copIteratorWorker.

1. For OLTP SQLs, it would be better build first 1 or 2 tasks in the same goroutine, then create a new goroutine to build the rest.
2. We can fix this issue when the request need an asending order firstly.
3. If the request need a desending order response, it may takes extra works on this issue. (We may need implement a desending spliting range method)

Difficulty

• Medium

Score

• 2500

Mentor(s)

• @SunRunAway
• @XuHuaiyu
• @qw4990

Recommended Skills

• Go programming

[ekalinin]

/pick-up-challenge

[sre-bot]

@ekalinin already had picked pingcap/tidb#12939, finish it before pick up a new one.

[ekalinin]

/pick-up-challenge

[sre-bot]

@ekalinin pick up issue success

[SunRunAway]

Hi @ekalinin , Thank you for picking this chanllenge. Do you have any ideas or plans? Feel free to ask me anything here or on slack.

[ekalinin]

Hi @ekalinin , Thank you for picking this chanllenge. Do you have any ideas or plans? Feel free to ask me anything here or on slack.

Hi @SunRunAway

Yes, i'm working on this issue it the moment. Current plan is:

• Add new function buildCopTasksChan. It's gonna be almost the same as buildCopTasks, except several things:
  • returns buffered channel (like: tasksCh := make(chan *copTask, 1024), 1024 is for example, we can think about more suitable value)
  • splitRanges / buildTiDBMemCopTasks will be called in background
  • splitRanges / buildTiDBMemCopTasks: all new instances of the copTask will be sent into that channel
• Remove buildCopTasks call from Send(...)
• Add buildCopTasksChan call during the copIterator.open(), returned channel will be saved in copIterator.tasksCh
• Replace copIteratorTaskSender.tasks []*copTask with copIteratorTaskSender.tasksCh chan *copTask,
• Update copIteratorTaskSender.run:
  • iterate over sender.tasksCh instead of sender.tasks
  • fill copIterator.tasks during iteration
• Update copIterator.Next:
  • for the keepOrder branch wait until all it.tasks are filled

Current state: did almost all steps, except last one (copIterator.Next). And a lot of tests are failing at the moment.

[SunRunAway]

👍 Sounds great, looking forward your pull request.

[SunRunAway]

@fzhedu PTAL when you're free, Thanks.

[you06]

After #53850, both tiflash and tikv coprocessor load the region info in batch from PD, building all the cop tasks no more taks a long time, I think this issue can be closed.