kyle/crossbar
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Wire wake and identity into crossbar; v2 smoke and README

Browse Source 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
This commit is contained in:
kyle
2026-09-25 11:12:24 -07:00
parent 47f49072cc
commit d0d3203f73
11 changed files with 300 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/smoke.sh
+40 -60
View File
@@ -1,80 +1,60 @@
#!/bin/sh
# Smoke run (v1): two fake upstreams, one crossbar with a fresh SQLite file, real HTTP.
# Checks routing, leases (sticky + header), failover, recovery, streaming, queueing, pin, drain,
# usage and metrics. Prints "smoke: ok" or fails with the crossbar log.
# Smoke run (v2): everything v1 checked, plus the context guard, wake-on-LAN and identity gating.
# 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 "s#^db .*#db = \"$tmp/crossbar.db\"#" example.toml > "$tmp/crossbar.toml"
bin/fakeupstream -listen 127.0.0.1:18081 -name alpha -models ornith-1.5-35b-a3b,small-9b -down-file "$tmp/alpha.down" -slow 600 >"$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" >"$tmp/beta.log" 2>&1 & pids="$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 1.5
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}' "$@"; }
# 1. a conversation gets a lease and keeps it; beta wins (2 slots × weight 2 vs 1 × 1)
# 1. v1 behaviour: with beta asleep, opencode-a goes to alpha
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv A)" "$base/opencode-a/v1/chat/completions")
[ "$h" = "200 beta new" ] || fail "first turn should be '200 beta new', got '$h'"
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv A)" "$base/opencode-a/v1/chat/completions")
[ "$h" = "200 beta reused" ] || fail "second turn should reuse beta, got '$h'"
[ "$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 $base/_crossbar/hosts)"
# 2. header route
h=$(hdrs -X POST -H 'Content-Type: application/json' -H 'X-Crossbar-Route: hermes-x' -d "$(conv B)" "$base/v1/chat/completions")
case "$h" in "200 beta new") ;; *) fail "header route hermes-x should be '200 beta new', got '$h'";; esac
h=$(curl -s -o /dev/null -w '%{http_code}' "$base/nope/v1/models"); [ "$h" = "404" ] || fail "unknown route 404, got $h"
# 2. context guard: a ~12k-token prompt does not fit alpha's 4096-token slot; beta is asleep and
# wakeable, so crossbar must send the magic packet, wait for beta, and place the prompt there.
start=$(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 "$base/_crossbar/hosts" | grep -q '"beta":{"healthy":true' || fail "beta not healthy after wake"
# 3. pin opencode-a to alpha: conversation A's next turn moves (an operator pin outranks the lease)
h=$(curl -s -o /dev/null -w '%{http_code}' -X POST -H 'Content-Type: application/json' -d '{"host":"alpha","pin":true}' "$base/_crossbar/routes/opencode-a")
[ "$h" = "200" ] || fail "pin returned $h"
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv A)" "$base/opencode-a/v1/chat/completions")
[ "$h" = "200 alpha new" ] || fail "after pin, conversation A should be '200 alpha new', got '$h'"
curl -s "$base/_crossbar/routes" | grep -q '"pinned":"alpha"' || fail "routes view does not show the pin: $(curl -s $base/_crossbar/routes)"
# 3. with beta awake, a prompt that fits nowhere is a 400 (both slots too small? no — beta fits):
# check the guard's refusal with a prompt beyond beta's 65536-per-slot too
# (the body goes through a file: a 300 KB string cannot be a single argv element on Linux)
{ printf '{"model":"ornith-1.5-35b-a3b","messages":[{"role":"user","content":"'; head -c 300000 /dev/zero | tr '\0' 'x'; printf '"}]}'; } > "$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. queue: alpha has parallel 1, queue_max 1, and answers in 600 ms → of three concurrent, one is 503
for i in 1 2 3; do (curl -s -o /dev/null -w '%{http_code}\n' -X POST -H 'Content-Type: application/json' -d "$(conv Q$i)" "$base/opencode-a/v1/chat/completions" >> "$tmp/codes") & sleep 0.1; done; wait $! 2>/dev/null || true
sleep 2.5
sort "$tmp/codes" | uniq -c | tr -s ' ' > "$tmp/counts"
grep -q '2 200' "$tmp/counts" && grep -q '1 503' "$tmp/counts" || fail "queue test wanted two 200 and one 503, got: $(cat "$tmp/counts")"
# 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 $h"
# 5. release the pin, drain alpha: new conversations go to beta, A stays on alpha
curl -s -o /dev/null -X POST -H 'Content-Type: application/json' -d '{"release":true}' "$base/_crossbar/routes/opencode-a"
h=$(curl -s -o /dev/null -w '%{http_code}' -X POST -H 'Content-Type: application/json' -d '{"drain":true}' "$base/_crossbar/hosts/alpha"); [ "$h" = "200" ] || fail "drain returned $h"
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv C)" "$base/opencode-a/v1/chat/completions")
[ "$h" = "200 beta new" ] || fail "with alpha draining a new conversation should go to beta, got '$h'"
curl -s "$base/_crossbar/hosts" | grep -q '"alpha":{[^}]*"draining":true' || fail "hosts view does not show alpha draining"
curl -s -o /dev/null -X POST -H 'Content-Type: application/json' -d '{"drain":false}' "$base/_crossbar/hosts/alpha"
# 6. failover + recovery
touch "$tmp/beta.down"; sleep 2.5
h=$(hdrs -X POST -H 'Content-Type: application/json' -d "$(conv C)" "$base/opencode-a/v1/chat/completions")
[ "$h" = "200 alpha new" ] || fail "with beta down conversation C should move to alpha, got '$h'"
curl -s "$base/_crossbar/hosts" | grep -q '"beta":{"healthy":false' || fail "hosts view does not show beta unhealthy"
rm "$tmp/beta.down"; sleep 3.5
curl -s "$base/_crossbar/hosts" | grep -q '"beta":{"healthy":true' || fail "beta did not recover after two good polls"
# 7. streaming still arrives incrementally, and the final usage chunk is untouched
# 5. v1 regression: streaming still incremental, usage and metrics present
start=$(date +%s%N)
curl -sN -X POST -H 'Content-Type: application/json' -d '{"model":"ornith-1.5-35b-a3b","stream":true,"messages":[{"role":"user","content":"stream me"}]}' \
"$base/opencode-a/v1/chat/completions" | while IFS= read -r line; do
[ -n "$line" ] || continue
now=$(date +%s%N); echo "$(( (now - start) / 1000000 )) $line"
done > "$tmp/stream.txt"
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: $(cat "$tmp/stream.txt")"
grep -q '"usage"' "$tmp/stream.txt" && grep -q 'DONE' "$tmp/stream.txt" || fail "stream lost the usage chunk or DONE"
# 8. accounting and metrics
[ -n "$firstms" ] && [ "$((lastms - firstms))" -ge 600 ] || fail "stream arrived in one burst"
sleep 1
u=$(curl -s "$base/_crossbar/usage?by=host")
echo "$u" | grep -q '"key":"alpha"' && echo "$u" | grep -q '"key":"beta"' || fail "usage by host: $u"
echo "$u" | grep -q '"cached_tokens":[1-9]' || fail "usage has no cached tokens (SSE/JSON usage not captured): $u"
curl -s -H 'Accept: text/plain' "$base/_crossbar/usage?by=route" | grep -qi 'cache' || fail "text usage table missing"
m=$(curl -s "$base/_crossbar/metrics")
echo "$m" | grep -q 'crossbar_requests_total{route="opencode-a",host="alpha",status="503"} 1' || fail "metrics missing the 503: $m"
echo "$m" | grep -q 'crossbar_host_healthy{host="beta"} 1' || fail "metrics missing host health"
grep -q 'route=opencode-a host=' "$tmp/crossbar.log" || fail "no request log line"
curl -s "$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"
echo "smoke: ok (stream spread $((lastms - firstms)) ms)"