[CORE-14786] rptest: speedup data_stat

[WillemKauf]

Currently, we use:

f"find {RedpandaService.DATA_DIR} -type f -exec stat -c '%n %s' '{{}}' \\;"

as a way to stat all of the files in a redpanda directory on a node.
This is bad: with -exec stat ... \;, find runs one stat process per file.

We can do better simply by changing this command instead to:

f"find {RedpandaService.DATA_DIR} -type f -exec stat -c '%n %s' '{{}}' +"

Where find will batch as many files as possible to each stat call.
However, we can do even better by not using -exec at all:

f"find {RedpandaService.DATA_DIR} -ignore_readdir_race -type f -printf '%p %s\\n'"

Here, we don't fork any external processes at all, and everything is handled natively in find. Failures due to the concurrent removal of files is silenced with the flag -ignore_readdir_race.

This should fix timeouts in CI for tests with large numbers of files, where 20+ minutes is spent waiting on these stat commands.

Backports Required

• none - not a bug fix
• none - this is a backport
• none - issue does not exist in previous branches
• none - papercut/not impactful enough to backport
• v26.1.x
• v25.3.x
• v25.2.x

Release Notes

• none

[WillemKauf]

For example, comparing the three options in my redpanda/src/v directory:

willem@bloom:~/redpanda/src/v$ time find -type f -exec stat -c '%n %s' '{}' \; > /dev/null

real	0m2.264s
user	0m1.287s
sys	0m0.972s
willem@bloom:~/redpanda/src/v$ time find  -type f -exec stat -c '%n %s' '{}' + > /dev/null

real	0m0.013s
user	0m0.001s
sys	0m0.011s
willem@bloom:~/redpanda/src/v$ time find -type f -printf '%p %s\n' > /dev/null

real	0m0.011s
user	0m0.001s
sys	0m0.010s

And within a directory with 1M files in 100 sub directories,

willem@bloom:/tmp/1m$ time find -type f -exec stat -c '%n %s' '{}' \; > /dev/null

real	9m47.619s
user	5m12.542s
sys	4m33.323s
willem@bloom:/tmp/1m$ time find -type f -exec stat -c '%n %s' '{}' + > /dev/null

real	0m1.571s
user	0m0.302s
sys	0m1.262s
willem@bloom:/tmp/1m$ time find -type f -printf '%p %s\n' > /dev/null

real	0m1.014s
user	0m0.191s
sys	0m0.819s

[Copilot]

Pull request overview

This PR optimizes RedpandaService.data_stat() in the rptest harness by avoiding per-file stat subprocess invocations, aiming to prevent CI timeouts when many files exist under the Redpanda data directory.

Changes:

• Replace find ... -exec stat ... \; with a find ... -printf ... approach to avoid forking stat for each file.
• Update inline commentary to reflect how transient filesystem races appear in output.

[vbotbuildovich]

Retry command for Build#83784

please wait until all jobs are finished before running the slash command

/ci-repeat 1
skip-redpanda-build
skip-units
skip-rebase
tests/rptest/tests/write_caching_fi_e2e_test.py::WriteCachingFailureInjectionE2ETest.test_crash_all@{"use_transactions":false}

[vbotbuildovich]

CI test results

test results on build#83784

test_status	test_class	test_method	test_arguments	test_kind	job_url	passed	reason	test_history
FLAKY(FAIL)	WriteCachingFailureInjectionE2ETest	test_crash_all	{"use_transactions": false}	integration	https://buildkite.com/redpanda/redpanda/builds/83784#019dd62d-95fe-4808-950f-f54e5ad1a4f7	14/21	Test FAILS after retries.Significant increase in flaky rate(baseline=0.0941, p0=0.0084, reject_threshold=0.0100)	https://redpanda.metabaseapp.com/dashboard/87-tests?tab=142-dt-individual-test-history&test_class=WriteCachingFailureInjectionE2ETest&test_method=test_crash_all

[oleiman]

🥳

[nvartolomei]

please backport to 26.1 and 25.3 to which we still frequently backport changes

[WillemKauf]

please backport to 26.1 and 25.3 to which we still frequently backport changes

Ack!

[vbotbuildovich]

/backport v26.1.x

[vbotbuildovich]

/backport v25.3.x