feat(gantry): improve cold-start distribution and benchmark automation - #586

Merged

Vaibhav Patel (vpatelsj) merged 62 commits into

mainfrom

wip/gantry-benchmark-observability

Aug 7, 2026

Commits on Aug 5, 2026

feat(gantry): capture benchmark performance telemetry
Vaibhav Patel
committed
ce116ed
feat(gantry): support private benchmark deployment
Vaibhav Patel
committed
2b4ff23
fix(gantry): isolate operator benchmark environment
Vaibhav Patel
committed
f3a0a25
fix(gantry): preserve detected operator context
Vaibhav Patel
committed
65ab22a
fix(gantry): shorten containerd metrics port name
Vaibhav Patel
committed
372ba06
fix(gantry): avoid AKS node exporter port conflict
Vaibhav Patel
committed
b4e7ad3
fix(gantry): reuse prepared images after preflight failure
Vaibhav Patel
committed
4a730ef
fix(gantry): check containerd build counter name
Vaibhav Patel
committed
ed8b039
feat(gantry): add repeatable benchmark deployment
Vaibhav Patel
committed
614af26
fix(gantry): open ACR before artifact lookup
Vaibhav Patel
committed
404a686
fix(gantry): publish artifacts without ACR lookup
Vaibhav Patel
committed
ebabed0
refactor(gantry): build deployment images privately
Vaibhav Patel
committed
2a70cb5
fix(gantry): wait for ACR task access
Vaibhav Patel
committed
c92e79d
fix(gantry): wait quietly for ACR task propagation
Vaibhav Patel
committed
fb97d78
fix(gantry): open ACR firewall for source build
Vaibhav Patel
committed
02ecb4d
fix(gantry): parse ACR readiness as JSON
Vaibhav Patel
committed
777d178
fix(gantry): ensure ACR private DNS zone groups
Vaibhav Patel
committed
dffb541
fix(gantry): serialize operator VM commands
Vaibhav Patel
committed
9a58b98
fix(gantry): validate operator disk size
Vaibhav Patel
committed
4b84462
fix(gantry): recover isolated private pull failures
Vaibhav Patel
committed
6266dc0
fix(gantry): reimage persistent private pull failures
Vaibhav Patel
committed
46bf1d7
fix(gantry): replace persistent private pull failures
Vaibhav Patel
committed
7cb5375
fix(gantry): preserve benchmark start request
Vaibhav Patel
committed
29f6a21
feat(gantry): show detailed benchmark progress
Vaibhav Patel
committed
f37910d
fix(gantry): bound containerd telemetry capture
Vaibhav Patel
committed
1fe1a51
feat(gantry): adopt retained benchmark images
Vaibhav Patel
committed
a454ffc

Commits on Aug 6, 2026

fix(gantry): wait for monitoring scrape coverage
Vaibhav Patel
committed
17443dc
style(gantry-benchmark): satisfy lint gates on benchmark and metrics code
Vaibhav Patel
committed
8d841ca
docs(gantry-benchmark): record Canada Central 1000-node results
Vaibhav Patel
committed
7492b16
fix(gantry-benchmark): drop max_concurrent_downloads override
Vaibhav Patel
committed
631c6b8
docs(gantry-benchmark): drop removed download concurrency note
Vaibhav Patel
committed
08a40d7
feat(gantry-benchmark): stream live image build and push progress
Vaibhav Patel
committed
97cbbdd
feat(gantry-benchmark): stream pull job pod progress
Vaibhav Patel
committed
b6a22e6
feat(gantry-benchmark): timestamp all progress output
Vaibhav Patel
committed
dbe8948
docs(gantry-benchmark): record Canada Central rerun results
Vaibhav Patel
committed
29ef0d0
feat(gantry-benchmark): restore max_concurrent_downloads override
Vaibhav Patel
committed
66ad0ed
docs(gantry-benchmark): analyze pod startup latency breakdown
Vaibhav Patel
committed
53c478e
docs(gantry-benchmark): attribute the Gantry gap to cold start
Vaibhav Patel
committed
9af85cd
docs(gantry-benchmark): explain the ramp as uniform layer ordering
Vaibhav Patel
committed
011a971
fix(gantry): run cold-start prefetch under live stream-through
Vaibhav Patel
committed
647eb4e

Commits on Aug 7, 2026

fix(gantry): resolve coordination peers to pod addresses
Vaibhav Patel
committed
8ac605b
fix(gantry): smooth redundant prefetch dispatch
Vaibhav Patel
committed
fab5bea
feat(gantry-benchmark): add live pull monitor
Vaibhav Patel
committed
0229391
feat(gantry-benchmark): show live pod states in monitor
Vaibhav Patel
committed
e8b41ac
fix(gantry): raise peer fetch ceiling to 15 minutes
Vaibhav Patel
committed
f8a520d
fix(gantry): bound cold-start coordination storms
Vaibhav Patel
committed
dff502d
feat(gantry): expose per-layer download progress
Vaibhav Patel
committed
da2d3c5
feat(benchmark): show per-node image progress
Vaibhav Patel
committed
5f34a23
feat(benchmark): prebuild reusable Gantry images
Vaibhav Patel
committed
864a4c4
fix(benchmark): start unloaded pool service
Vaibhav Patel
committed
37a4583
feat(benchmark): report pool cleanup status
Vaibhav Patel
committed
58c765b
feat(benchmark): monitor per-node resources
Vaibhav Patel
committed
66408e1
feat(gantry): add diagnostic CPU profiling
Vaibhav Patel
committed
c622c1a
perf(gantry): batch peer stream copies
Vaibhav Patel
committed
be60813
perf(gantry): enlarge peer HTTP2 frames
Vaibhav Patel
committed
4c6d587
fix(benchmark): preserve partial CPU profiles
Vaibhav Patel
committed
03bc52f
feat(benchmark): add fresh Gantry run wrapper
Vaibhav Patel
committed
de9f6be
fix: address Gantry review findings
Vaibhav Patel
committed
7d82cb5
fix(benchmark): guard node header index
Vaibhav Patel
committed
9fe97b7
Merge branch 'main' into wip/gantry-benchmark-observability
Vaibhav Patel (vpatelsj)
authored
9d87055
chore: remove Gantry benchmark binary
Vaibhav Patel
committed
f3426c0
Merge branch 'main' into wip/gantry-benchmark-observability
Vaibhav Patel (vpatelsj)
authored
a31d753