pd-simulator: update import-data case #2741
Conversation
| @@ -40,8 +40,8 @@ func newImportData() *Case { | |||
| }) | |||
| } | |||
|
|
|||
| storeIDs := rand.Perm(3) | |||
| for i := 0; i < 40; i++ { | |||
| for i := 0; i < 10000; i++ { | |||
There was a problem hiding this comment.
Changed to configurable.
Which one... |
|
PTAL @nolouch |
|
| checkCount++ | ||
| dev := 0.0 | ||
| for _, p := range regionProps { | ||
| dev += (p - 10) * (p - 10) / 100 |
There was a problem hiding this comment.
will it still be 10 after region num is configured?
There was a problem hiding this comment.
It means if balanced, store each get 10 % of data, no matter what the total is.
In fact, after changing to the new scenario, because import data(3 out of 10 peers) makes it less balanced, we don't need to set a stop time, but we should observe the deviation. If schedulers work properly, balance should not be greatly affected by the import. |
This behavior seems to be different compared to other simulators |
Yes. It is because the new scenario is an unbalanced import. Indeed, it is trivial to add a stop condition. |
It will stop at tick 1009 if region number is 10000. |
|
/merge |
|
/merge |
ti-srebot
added
the
status/can-merge
|
/run-all-tests |
* Fix mv peer log parse error * Delete unused var * Rename * Update case * Address comment * Update import_data.go * Add stop condition Co-authored-by: ShuNing <nolouch@gmail.com>
* Fix mv peer log parse error * Delete unused var * Rename * Update case * Address comment * Update import_data.go * Add stop condition Co-authored-by: ShuNing <nolouch@gmail.com> Signed-off-by: ZenoTan <zenotan1998@gmail.com>
What problem does this PR solve?
Import data case is a case in pd simulator. Originally, it has 40 regions, and it imports many regions afterwards. Then import regions >> previous regions. The real scenario is, import regions << previous regions. So now it imports regions 10% of previous size.
What is changed and how it works?
Changed import data case.
Check List
Tests
Release note