tests: use admission.io.overload in admission-control/elastic-io

[sumeerbhola]

The previously used sub-level metric was flawed, in that the IO overload score could stay low even at higher sub-level counts if L0 had very few bytes (which is a deliberate choice in admission control). So admission control would not throttle elastic work as aggressively as the test expected it to.

Running with this change, we don't exceed a score of 0.15, while the previously used sub-level count metric spikes higher. For example:

2025/10/14 22:13:15 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.100000
2025/10/14 22:13:25 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.100000

I251014 22:13:27.665296 868 util/admission/io_load_listener.go:780 ⋮ [T1,Vsystem,n1,s1] 2918  IO overload: compaction score 0.150 (131 ssts, 9 sub-levels), L0 growth 551 MiB (write 551 MiB (ignored 0 B) ingest 0 B (ignored 0 B)): requests 15985 (0 bypassed) with 505 MiB acc-write (0 B bypassed) + 0 B acc-ingest (0 B bypassed) + 551 MiB adjusted-LSM-writes + 4.2 GiB adjusted-disk-writes + write-model 1.09x+1 B (smoothed 1.08x+1 B) + l0-ingest-model 0.00x+0 B (smoothed 0.75x+1 B) + ingest-model 0.00x+0 B (smoothed 1.00x+1 B) + write-amp-model 7.87x+1 B (smoothed 8.01x+1 B) + at-admission-tokens 126 B, compacted 550 MiB [≈545 MiB], flushed 799 MiB [≈838 MiB] (mult 0.77); admitting 649 MiB (rate 43 MiB/s) (elastic 519 MiB rate 35 MiB/s) due to memtable flush (multiplier 0.775) (used total: 543 MiB elastic 541 MiB); write stalls 0; diskBandwidthLimiter (unlimited) (tokenUtilization 0.00, tokensUsed (elastic 0 B, snapshot 0 B, regular 0 B) tokens (write 0 B (prev 0 B), read 0 B (prev 0 B)), writeBW 0 B/s, readBW 0 B/s, provisioned 0 B/s)

2025/10/14 22:13:35 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.050000

I251014 22:13:42.666326 868 util/admission/io_load_listener.go:780 ⋮ [T1,Vsystem,n1,s1] 2926  IO overload: compaction score 0.050 (70 ssts, 5 sub-levels), L0 growth 498 MiB (write 498 MiB (ignored 0 B) ingest 0 B (ignored 0 B)): requests 15228 (0 bypassed) with 480 MiB acc-write (0 B bypassed) + 0 B acc-ingest (0 B bypassed) + 498 MiB adjusted-LSM-writes + 4.2 GiB adjusted-disk-writes + write-model 1.04x+1 B (smoothed 1.06x+1 B) + l0-ingest-model 0.00x+0 B (smoothed 0.75x+1 B) + ingest-model 0.00x+0 B (smoothed 1.00x+1 B) + write-amp-model 8.57x+1 B (smoothed 8.29x+1 B) + at-admission-tokens 153 B, compacted 498 MiB [≈522 MiB], flushed 883 MiB [≈860 MiB] (mult 0.77); admitting 667 MiB (rate 44 MiB/s) (elastic 533 MiB rate 36 MiB/s) due to memtable flush (multiplier 0.775) (used total: 519 MiB elastic 517 MiB); write stalls 0; diskBandwidthLimiter (unlimited) (tokenUtilization 0.00, tokensUsed (elastic 0 B, snapshot 0 B, regular 0 B) tokens (write 0 B (prev 0 B), read 0 B (prev 0 B)), writeBW 0 B/s, readBW 0 B/s, provisioned 0 B/s)

Fixes #148786
Fixes #156168
Fixes #156215

Epic: none

Release note: None

[cockroach-teamcity]

This change is 

[tbg]

Thank you!

bors r+

[craig]

Build failed (retrying...):

• examples_orms

[craig]

Build succeeded:

• linux_amd64_fips_build
• docker_image_amd64
• examples_orms
• acceptance
• local_roachtest_fips
• lint
• linux_amd64_build
• check_generated_code
• unit_tests
• local_roachtest

[blathers-crl]

Based on the specified backports for this PR, I applied new labels to the following linked issue(s). Please adjust the labels as needed to match the branches actually affected by the issue(s), including adding any known older branches.

---

Issue #156215: branch-release-25.4.

---

Issue #148786: branch-release-25.4.

---

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}