72 lines
6.4 KiB
Bash
72 lines
6.4 KiB
Bash
#!/bin/sh
|
|
# Smoke run (v2.3): everything v1 checked, plus the context guard, wake-on-LAN, identity gating and
|
|
# a route's dedicated listener.
|
|
# Prints "smoke: ok" or fails with the crossbar log.
|
|
set -eu
|
|
cd "$(dirname "$0")/.."
|
|
tmp=$(mktemp -d); trap 'kill $pids 2>/dev/null; rm -rf "$tmp"' EXIT INT TERM
|
|
pids=""
|
|
sed -e "s#^db .*#db = \"$tmp/crossbar.db\"#" -e 's#^identity .*#identity = "header"#' -e 's#^\# peers = \["talos"\]#peers = ["talos"]#' example.toml > "$tmp/crossbar.toml"
|
|
# alpha: small context (4096 per slot = 8192/2); beta: large, sleeps until woken
|
|
bin/fakeupstream -listen 127.0.0.1:18081 -name alpha -models ornith-1.5-35b-a3b,small-9b -down-file "$tmp/alpha.down" -slow 600 -n-ctx 8192 -slots 2 >"$tmp/alpha.log" 2>&1 & pids="$pids $!"
|
|
bin/fakeupstream -listen 127.0.0.1:18082 -name beta -models ornith-1.5-35b-a3b -down-file "$tmp/beta.down" -n-ctx 131072 -slots 2 -wol-listen 127.0.0.1:19082 -wol-mac aa:bb:cc:dd:ee:02 >"$tmp/beta.log" 2>&1 & pids="$pids $!"
|
|
touch "$tmp/beta.down" # beta starts "asleep"
|
|
bin/crossbar -config "$tmp/crossbar.toml" >"$tmp/crossbar.log" 2>&1 & pids="$pids $!"
|
|
sleep 2.5 # two polls: alpha healthy, beta down
|
|
fail() { echo "smoke: FAIL: $*" >&2; echo "--- crossbar.log"; cat "$tmp/crossbar.log"; exit 1; }
|
|
base=http://127.0.0.1:17777
|
|
conv() { printf '{"model":"ornith-1.5-35b-a3b","stream":false,"messages":[{"role":"system","content":"smoke"},{"role":"user","content":"conversation %s"}]}' "$1"; }
|
|
big() { printf '{"model":"ornith-1.5-35b-a3b","stream":false,"messages":[{"role":"user","content":"%s"}]}' "$(head -c 40000 /dev/zero | tr '\0' 'x')"; }
|
|
hdrs() { curl -s -o /dev/null -w '%{http_code} %header{X-Crossbar-Host} %header{X-Crossbar-Lease}' "$@"; }
|
|
|
|
# 1. v1 behaviour: with beta asleep, opencode-a goes to alpha
|
|
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv A)" "$base/opencode-a/v1/chat/completions")
|
|
[ "$h" = "200 alpha new" ] || fail "with beta asleep conversation A should be '200 alpha new', got '$h'"
|
|
curl -s "$base/_crossbar/hosts" | grep -q '"alpha":{[^}]*"n_ctx":8192' || fail "hosts view does not show alpha n_ctx 8192: $(curl -s $base/_crossbar/hosts)"
|
|
|
|
# 2. context guard: a ~12k-token prompt does not fit alpha's 4096-token slot; beta is asleep and
|
|
# wakeable, so crossbar must send the magic packet, wait for beta, and place the prompt there.
|
|
start=$(date +%s)
|
|
h=$(hdrs -m 40 -X POST -H 'Content-Type: application/json' -d "$(big)" "$base/opencode-a/v1/chat/completions")
|
|
[ "$h" = "200 beta new" ] || fail "oversized prompt should wake beta and land there, got '$h' after $(( $(date +%s) - start ))s"
|
|
grep -q "magic packet received" "$tmp/beta.log" || fail "beta never saw a magic packet"
|
|
curl -s "$base/_crossbar/hosts" | grep -q '"beta":{"healthy":true' || fail "beta not healthy after wake"
|
|
|
|
# 3. with beta awake, a prompt that fits nowhere is a 400 (both slots too small? no — beta fits):
|
|
# check the guard's refusal with a prompt beyond beta's 65536-per-slot too
|
|
# (the body goes through a file: a 300 KB string cannot be a single argv element on Linux)
|
|
{ printf '{"model":"ornith-1.5-35b-a3b","messages":[{"role":"user","content":"'; head -c 300000 /dev/zero | tr '\0' 'x'; printf '"}]}'; } > "$tmp/toolarge-req.json"
|
|
h=$(curl -s -o "$tmp/toolarge.json" -w '%{http_code}' -X POST -H 'Content-Type: application/json' -d @"$tmp/toolarge-req.json" "$base/opencode-a/v1/chat/completions")
|
|
[ "$h" = "400" ] && grep -q '"prompt too large"' "$tmp/toolarge.json" || fail "300 KB prompt should be 400 prompt too large, got $h $(cat "$tmp/toolarge.json")"
|
|
|
|
# 4. identity: hermes-x is locked to peer talos (header mode)
|
|
h=$(curl -s -o /dev/null -w '%{http_code}' -X POST -H 'Content-Type: application/json' -d "$(conv B)" "$base/hermes-x/v1/chat/completions")
|
|
[ "$h" = "403" ] || fail "hermes-x without a peer header should be 403, got $h"
|
|
h=$(curl -s -o /dev/null -w '%{http_code}' -X POST -H 'Content-Type: application/json' -H 'X-Crossbar-Peer: titan' -d "$(conv B)" "$base/hermes-x/v1/chat/completions")
|
|
[ "$h" = "403" ] || fail "hermes-x as titan should be 403, got $h"
|
|
h=$(hdrs -X POST -H 'Content-Type: application/json' -H 'X-Crossbar-Peer: talos' -d "$(conv B)" "$base/hermes-x/v1/chat/completions")
|
|
case "$h" in 200*) ;; *) fail "hermes-x as talos should be 200, got '$h'";; esac
|
|
h=$(curl -s -o /dev/null -w '%{http_code}' "$base/_crossbar/hosts"); [ "$h" = "200" ] || fail "admin must not be gated, got $h"
|
|
|
|
# 5. v1 regression: streaming still incremental, usage and metrics present
|
|
start=$(date +%s%N)
|
|
curl -sN -X POST -H 'Content-Type: application/json' -H 'X-Crossbar-Peer: talos' -d '{"model":"ornith-1.5-35b-a3b","stream":true,"messages":[{"role":"user","content":"stream me"}]}' \
|
|
"$base/hermes-x/v1/chat/completions" | while IFS= read -r line; do [ -n "$line" ] || continue; now=$(date +%s%N); echo "$(( (now - start) / 1000000 )) $line"; done > "$tmp/stream.txt"
|
|
firstms=$(head -1 "$tmp/stream.txt" | cut -d' ' -f1); lastms=$(tail -1 "$tmp/stream.txt" | cut -d' ' -f1)
|
|
[ -n "$firstms" ] && [ "$((lastms - firstms))" -ge 600 ] || fail "stream arrived in one burst"
|
|
sleep 1
|
|
curl -s "$base/_crossbar/usage?by=host" | grep -q '"cached_tokens":[1-9]' || fail "usage has no cached tokens"
|
|
curl -s "$base/_crossbar/metrics" | grep -q 'crossbar_requests_total{route="hermes-x",host="beta",status="403"}' && fail "403s are refused before a lease and must not be counted as requests"
|
|
curl -s "$base/_crossbar/metrics" | grep -q 'crossbar_host_healthy{host="beta"} 1' || fail "metrics missing beta health"
|
|
# 6. v2.3: boxmaker-a has its own listener. Paths are unprefixed, the chat and a control call land
|
|
# on the same host (affinity = "route"), and the admin API is not served there.
|
|
lb=http://127.0.0.1:17801
|
|
h1=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv C)" "$lb/v1/chat/completions")
|
|
h2=$(hdrs "$lb/props?model=ornith-1.5-35b-a3b")
|
|
case "$h1" in 200*) ;; *) fail "chat on the dedicated listener should be 200, got '$h1'";; esac
|
|
[ "$(echo "$h1" | cut -d' ' -f2)" = "$(echo "$h2" | cut -d' ' -f2)" ] || fail "chat went to '$h1', /props to '$h2': one route, one host"
|
|
h=$(curl -s -o /dev/null -w '%{http_code}' "$lb/_crossbar/hosts"); [ "$h" = "404" ] || fail "admin must not be served on a dedicated listener, got $h"
|
|
h=$(curl -s -o /dev/null -w '%{http_code}' "$lb/boxmaker-a/v1/models"); [ "$h" = "404" ] || fail "a prefixed path on the dedicated listener should be 404, got $h"
|
|
|
|
echo "smoke: ok (stream spread $((lastms - firstms)) ms)"
|