enable but inact memory profiling

[hicqu]

Signed-off-by: qupeng qupeng@pingcap.com

What problem does this PR solve?

Close #1270 and tikv/tikv#9538.

Check List

Tests

• Manual test (add detailed scripts or steps below)
  1. Start a TiKV cluster with the environment variable set.
  2. wget "http://172.16.4.78:50013/debug/pprof/heap?seconds=1" to dump a memory profiling

Code changes

Side effects

Related changes

• Need to cherry-pick to the release branch
• Need to update the documentation

Release notes:

NONE

[CLAassistant]

All committers have signed the CLA.

[lucklove]

/lgtm

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• lucklove

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by writing /lgtm in a comment.
Reviewer can cancel approval by writing /lgtm cancel in a comment.

[lucklove]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: d65016d

[lucklove]

/hold

[lucklove]

To fix tikv/tikv#9538, maybe you want to change tiup-playground instead of tiup-cluster?

[hicqu]

To fix tikv/tikv#9538, maybe you want to change tiup-playground instead of tiup-cluster?

@lucklove all TiKV clusters have the problem, regardless they are deployed with playground or cluster.

This PR is only for cluster, do you have any idea for setting environment variables for playground?

[hicqu]

BTW I will do a bench for this change. The result will be post later.

[lucklove]

To fix tikv/tikv#9538, maybe you want to change tiup-playground instead of tiup-cluster?

@lucklove all TiKV clusters have the problem, regardless they are deployed with playground or cluster.

This PR is only for cluster, do you have any idea for setting environment variables for playground?

TiUP set envs for each component here.

1. You can add a Envs []string filed for PrepareCommandParams
2. And add envs = append(envs, p.Envs) on L224.
3. Then add envs arg for NewComponentProcess
4. Add your env when call NewComponentProcess in TiKVInstance

[codecov-io]

Codecov Report

Merging #1272 (d28621f) into master (c9b58c9) will increase coverage by 27.37%.
The diff coverage is n/a.

❗ Current head d28621f differs from pull request most recent head d65016d. Consider uploading reports for the commit d65016d to get more accurate results

@@             Coverage Diff             @@
##           master    #1272       +/-   ##
===========================================
+ Coverage   26.28%   53.66%   +27.37%     
===========================================
  Files         263      289       +26     
  Lines       18840    20780     +1940     
===========================================
+ Hits         4953    11152     +6199     
+ Misses      13115     7891     -5224     
- Partials      772     1737      +965     

Flag	Coverage Δ
cluster	45.14% <ø> (?)
dm	26.70% <ø> (?)
integrate	48.16% <ø> (+31.62%)	⬆️
playground	3.11% <ø> (?)
tiup	16.56% <ø> (+0.02%)	⬆️
unittest	22.81% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pkg/logger/logger.go	100.00% <0.00%> (ø)
components/playground/instance/ticdc.go	0.00% <0.00%> (ø)
pkg/cluster/operation/scale_in.go	58.27% <0.00%> (ø)
components/playground/instance/process.go	0.00% <0.00%> (ø)
pkg/logger/debug.go	16.66% <0.00%> (ø)
pkg/cluster/operation/check.go	47.76% <0.00%> (ø)
pkg/logger/audit.go	53.33% <0.00%> (ø)
pkg/cluster/operation/telemetry.go	0.00% <0.00%> (ø)
components/playground/instance/tiflash.go	0.00% <0.00%> (ø)
pkg/cluster/operation/action.go	60.63% <0.00%> (ø)
... and 203 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c9b58c9...d65016d. Read the comment docs.

[hicqu]

I have benchmarked it, here is the result:

without any environment variables:
$ cargo bench
running 7 tests
test bench_malloc_128k ... bench:          39 ns/iter (+/- 7)
test bench_malloc_128m ... bench:       4,134 ns/iter (+/- 412)
test bench_malloc_1g   ... bench:       8,478 ns/iter (+/- 2,356)
test bench_malloc_1k   ... bench:          48 ns/iter (+/- 5)
test bench_malloc_1m   ... bench:          37 ns/iter (+/- 5)
test bench_malloc_8k   ... bench:          30 ns/iter (+/- 4)
test bench_malloc_8m   ... bench:          35 ns/iter (+/- 3)

with `export MALLOC_CONF="prof:true,prof_active:true,prof.active:false"`
$ cargo bench
running 7 tests
test bench_malloc_128k ... bench:          37 ns/iter (+/- 16)
test bench_malloc_128m ... bench:       4,136 ns/iter (+/- 605)
test bench_malloc_1g   ... bench:       8,426 ns/iter (+/- 1,134)
test bench_malloc_1k   ... bench:          44 ns/iter (+/- 6)
test bench_malloc_1m   ... bench:          35 ns/iter (+/- 3)
test bench_malloc_8k   ... bench:          32 ns/iter (+/- 3)
test bench_malloc_8m   ... bench:          37 ns/iter (+/- 5)

There aren't obvious dirrerences.

The reason is memory profiling is compiled for both 2 cases, and it's not actived in both 2 cases.

[lucklove]

/hold cancel

[lucklove]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: c8f2b81