Fence writers and recover stale Binance WebSockets #1

Merged
magnus merged 2 commits from fix/websocket-failover into master 2026-09-19 18:09:20 +00:00
12 changed files with 336 additions and 129 deletions
Showing only changes of commit 904f3fffe6 - Show all commits
+12
View File
@@ -0,0 +1,12 @@
name: pr
on:
pull_request:
branches: [master]
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Unit and PostgreSQL fencing tests
run: bash scripts/test-fencing.sh
# No registry, deployment, or production credentials in PR jobs.
+27
View File
@@ -1,5 +1,32 @@
# btcdata Kubernetes deployment plan
## 2026-09-19 status and proposed recovery
Kubernetes deployment already exists: one `Recreate` replica on citadel, eligible
on citadel and srv1893094. Both nodes were Ready. At 14:03 UTC the live price
WebSocket upgraded successfully but returned data last recorded at 08:36:42 UTC.
Harbor was refusing connections and Flux apps reconciliation was dependency-blocked.
No production restart or relocation was performed during this investigation.
The `fix/websocket-failover` application change proposes bounded Binance reconnects,
60-second in-memory freshness readiness, and PostgreSQL single-writer fencing.
Every application query uses the session holding advisory lock `728194106`; loss
of that session cannot cause writes through another pooled connection. Competing
instances fail startup rather than becoming simultaneous writers. This requires
all writers to run the fenced version; legacy deployments must stay stopped.
Companion deployment changes in `magnus/deploy` add a **suspended** watchdog:
three failed minute checks after startup grace request relocation to the other
healthy node, with a ten-minute cooldown. See `apps/btcdata/FAILOVER.md` in that
repository for review gates, acceptance tests, limitations, and rollback. This is
proposed code, not a claim that automatic failover is enabled in production.
Run `make test` for unit tests. Run `bash scripts/test-fencing.sh` for unit tests
plus PostgreSQL lock-contention, session-loss, and pooled-lock-release tests in
isolated disposable containers. No production credentials are required.
The remaining sections below are the original migration plan, not current status.
## Goal and interpretation
Deploy `btcdata` through the existing Flux-managed two-node k3s cluster (`srv1893094` and `citadel`), with **exactly one btcdata pod running at a time**. The pod may run on either node and Kubernetes may recreate it on the other node after failure.
+12 -7
View File
@@ -43,36 +43,41 @@
:db {:datasource #ig/ref :db.sql/connection}
:migrate-on-init? true}
:writer/query-fn
{:conn #ig/ref :db.sql/connection
:query-fn #ig/ref :db.sql/query-fn
:migrations #ig/ref :db.sql/migrations}
:reitit.routes/api
{:base-path "/api"
:query-fn #ig/ref :db.sql/query-fn
:query-fn #ig/ref :writer/query-fn
:binance #ig/ref :ws/binance
:frankfurter #ig/ref :frankfurter/rates
:strike #ig/ref :strike/ticker}
:ws/binance
{:query-fn #ig/ref :db.sql/query-fn
{:query-fn #ig/ref :writer/query-fn
:uri "wss://stream.binance.com:9443/ws/btcusdt@trade"
:enabled? #or [#env BINANCE_ENABLED "true"]}
:kraken/ohlc
{:query-fn #ig/ref :db.sql/query-fn
{:query-fn #ig/ref :writer/query-fn
:enabled? #or [#env KRAKEN_ENABLED "true"]}
:kraken/ohlc-day
{:query-fn #ig/ref :db.sql/query-fn
{:query-fn #ig/ref :writer/query-fn
:enabled? #or [#env KRAKEN_ENABLED "true"]}
:kraken/ohlc-minute
{:query-fn #ig/ref :db.sql/query-fn
{:query-fn #ig/ref :writer/query-fn
:enabled? #or [#env KRAKEN_ENABLED "true"]}
:frankfurter/rates
{:url #or [#env FRANKFURTER_URL "http://localhost:8080"]
:query-fn #ig/ref :db.sql/query-fn}
:query-fn #ig/ref :writer/query-fn}
:strike/ticker
{:query-fn #ig/ref :db.sql/query-fn
{:query-fn #ig/ref :writer/query-fn
:api-key #env STRIKE_API_KEY
:sats-eur-amount #or [#env SATS_EUR_AMOUNT "55"]
:enabled? #or [#env STRIKE_ENABLED "true"]}}
+27
View File
@@ -0,0 +1,27 @@
#!/usr/bin/env bash
# Disposable isolated PostgreSQL; no production credentials or published ports.
set -euo pipefail
name="btcdata-test-${GITHUB_RUN_ID:-$$}-${GITHUB_RUN_ATTEMPT:-1}"
cleanup() {
docker rm -f "$name" >/dev/null 2>&1 || true
docker network rm "$name" >/dev/null 2>&1 || true
}
trap cleanup EXIT
docker network create "$name" >/dev/null
docker run -d --name "$name" --network "$name" \
-e POSTGRES_HOST_AUTH_METHOD=trust postgres:16-alpine >/dev/null
ready=false
for _ in $(seq 1 60); do
if docker exec "$name" pg_isready -U postgres >/dev/null 2>&1; then
ready=true
break
fi
sleep 1
done
if [ "$ready" != true ]; then
echo 'Disposable PostgreSQL did not become ready' >&2
exit 1
fi
docker run --rm --network "$name" -v "$PWD:/app" -w /app \
-e "BTCDATA_TEST_JDBC_URL=jdbc:postgresql://$name:5432/postgres?user=postgres" \
clojure:temurin-21-tools-deps-alpine clj -M:test
+1
View File
@@ -16,6 +16,7 @@
[pmagnus.btcdata.web.routes.api]
;; WebSocket clients
[pmagnus.btcdata.ws.binance]
[pmagnus.btcdata.writer]
;; Pollers
[pmagnus.btcdata.kraken.ohlc]
[pmagnus.btcdata.kraken.ohlc-daily]
@@ -1,7 +1,8 @@
(ns pmagnus.btcdata.web.controllers.health
(:require
[clojure.tools.logging :as log]
[ring.util.http-response :as response])
[ring.util.http-response :as response]
[pmagnus.btcdata.ws.binance :as binance])
(:import
[java.lang.management ManagementFactory]))
@@ -12,11 +13,13 @@
:app {:status "up"}}))
(defn readiness-check!
"Report readiness only while the application can query PostgreSQL."
[query-fn _req]
"Require the lock-owning database session and a fresh upstream stream."
[query-fn component _req]
(try
(query-fn :check-database {})
(response/ok {:status "ready"})
(if (binance/fresh? component)
(response/ok {:status "ready"})
(response/service-unavailable {:status "not ready" :reason "binance stale"}))
(catch Exception e
(log/warn e "Database readiness check failed")
(response/service-unavailable {:status "not ready"}))))
+1 -1
View File
@@ -181,7 +181,7 @@
["/health"
{:get health/healthcheck!}]
["/ready"
{:get (fn [req] (health/readiness-check! (:query-fn opts) req))}]
{:get (fn [req] (health/readiness-check! (:query-fn opts) (:binance opts) req))}]
["/price/ws"
{:get (fn [req] (ws-price-handler opts req))
:no-doc true
+36
View File
@@ -0,0 +1,36 @@
(ns pmagnus.btcdata.writer
"Fence all application queries with one PostgreSQL session advisory lock.
Queries use the lock-owning connection itself: a lost session cannot keep
writing through a different pooled connection. Never reconnect this session."
(:require [integrant.core :as ig]
[next.jdbc :as jdbc]))
(def lock-id 728194106)
(defn guarded-query [connection query-fn]
(fn [query params]
(locking connection
(query-fn connection query params))))
(defmethod ig/init-key :writer/query-fn
[_ {:keys [conn query-fn]}]
(let [connection (jdbc/get-connection conn)]
(try
;; Bound lock/query stalls; a failed session is never replaced behind the fence.
(.setNetworkTimeout connection
(reify java.util.concurrent.Executor (execute [_ task] (.run ^Runnable task)))
10000)
(when-not (:locked (jdbc/execute-one! connection
["select pg_try_advisory_lock(?) as locked" lock-id]))
(throw (ex-info "Another btcdata writer holds the database lock" {})))
(with-meta (guarded-query connection query-fn) {:connection connection})
(catch Exception e
(.close connection)
(throw e)))))
(defmethod ig/halt-key! :writer/query-fn [_ query-fn]
(let [connection (:connection (meta query-fn))]
(locking connection
(try
(jdbc/execute-one! connection ["select pg_advisory_unlock(?)" lock-id])
(finally (.close connection))))))
+93 -88
View File
@@ -1,97 +1,102 @@
(ns pmagnus.btcdata.ws.binance
(:require
[clojure.data.json :as json]
[clojure.tools.logging :as log]
[integrant.core :as ig])
(:import
[java.net URI]
[java.net.http HttpClient WebSocket WebSocket$Listener]
[java.util.concurrent CompletableFuture CompletionStage]))
(:require [clojure.data.json :as json]
[clojure.tools.logging :as log]
[integrant.core :as ig])
(:import [java.net URI]
[java.net.http HttpClient WebSocket WebSocket$Listener]
[java.time Duration Instant]
[java.util.concurrent CompletableFuture TimeUnit]))
(defn- save-price! [query-fn price]
(try
(query-fn :insert-binance-price! {:price price})
(catch Exception e
(log/error e "Failed to save BTC price"))))
(def stale-after-ms 60000)
(defn- connect!
"Opens a WebSocket to Binance trade stream and returns the WebSocket instance.
`state` is an atom with keys :running?, :last-write, :buffer.
`latest-price` is an atom reset with each throttled price for SSE consumers."
[uri query-fn state latest-price]
(let [client (HttpClient/newHttpClient)
listener (reify WebSocket$Listener
(onOpen [_ ws]
(log/info "Binance WebSocket connected")
(.request ws 1))
(defn fresh?
"Only in-process stream data counts; database fallback is not stream health."
([component] (fresh? component (System/currentTimeMillis)))
([{:keys [state latest-price]} now]
(let [{:keys [enabled? connected?]} @state
recorded-at (:recorded-at @latest-price)]
(or (false? enabled?)
(and connected? (some? recorded-at)
(<= 0 (- now (.toEpochMilli ^Instant recorded-at)) stale-after-ms))))))
(onText [_ ws data last?]
(let [buf (:buffer @state)]
(.append buf data)
(when last?
(let [text (str buf)]
(.setLength buf 0)
(try
(let [msg (json/read-str text :key-fn keyword)
price (some-> (:p msg) bigdec)]
(when price
(let [now (System/currentTimeMillis)]
(when (> (- now (:last-write @state)) 5000)
(swap! state assoc :last-write now)
(log/info "BTC price:" (str price))
(save-price! query-fn price)
(let [prev-price (:price @latest-price)]
(reset! latest-price
{:price price
:prev-price prev-price
:recorded-at (java.time.Instant/now)}))))))
(catch Exception e
(log/error e "Failed to parse Binance message")))))
(let [^CompletionStage cf (CompletableFuture/completedFuture nil)]
(.request ws 1)
cf)))
(onClose [_ _ws code reason]
(log/warn "Binance WebSocket closed:" code reason)
(when (:running? @state)
(future
(Thread/sleep 3000)
(when (:running? @state)
(log/info "Reconnecting to Binance...")
(try
(let [ws (connect! uri query-fn state latest-price)]
(swap! state assoc :ws ws))
(catch Exception e
(log/error e "Binance reconnect failed")))))))
(onError [_ _ws error]
(log/error error "Binance WebSocket error")))]
(-> (.newWebSocketBuilder client)
(defn- connect! [client uri query-fn state latest-price generation]
(let [buffer (StringBuilder.)
current? #(and (:running? @state) (= generation (:generation @state)))
listener
(reify WebSocket$Listener
(onOpen [_ ws]
(if (current?)
(do (swap! state assoc :connected? true)
(log/info "Binance WebSocket connected")
(.request ws 1))
(.abort ws)))
(onText [_ ws data last?]
(when (current?)
(.append buffer data)
(when last?
(let [text (str buffer)]
(.setLength buffer 0)
(try
(when-let [price (some-> (json/read-str text :key-fn keyword) :p bigdec)]
(let [now (System/currentTimeMillis)]
(swap! state assoc :last-message now)
(when (> (- now (:last-write @state)) 5000)
;; Publish only after persistence succeeds.
(query-fn :insert-binance-price! {:price price})
(swap! state assoc :last-write now)
(reset! latest-price {:price price
:prev-price (:price @latest-price)
:recorded-at (Instant/now)}))))
(catch Exception e
(log/error e "Failed to process Binance price"))))))
(.request ws 1)
(CompletableFuture/completedFuture nil))
(onClose [_ _ code reason]
(when (current?)
(swap! state assoc :connected? false)
(log/warn "Binance WebSocket closed:" code reason))
(CompletableFuture/completedFuture nil))
(onError [_ _ error]
(when (current?)
(swap! state assoc :connected? false)
(log/warn error "Binance WebSocket error"))))]
(-> (.newWebSocketBuilder ^HttpClient client)
(.connectTimeout (Duration/ofSeconds 10))
(.buildAsync (URI. uri) listener)
(.join))))
(.get 15 TimeUnit/SECONDS))))
(defmethod ig/init-key :ws/binance
[_ {:keys [query-fn uri enabled?] :or {enabled? "true"}}]
(if-not (= enabled? "true")
(do (log/info "Binance WebSocket disabled")
{:state (atom {:running? false})
:latest-price (atom nil)})
(do (log/info "Starting Binance WebSocket listener:" uri)
(let [latest-price (atom nil)
state (atom {:running? true
:last-write 0
:buffer (StringBuilder.)
:ws nil})
ws (connect! uri query-fn state latest-price)]
(swap! state assoc :ws ws)
{:state state
:latest-price latest-price}))))
(let [enabled (= enabled? "true")
latest-price (atom nil)
state (atom {:enabled? enabled :running? enabled :connected? false
:last-write 0 :last-message 0 :generation nil :ws nil})
client (when enabled (-> (HttpClient/newBuilder)
(.connectTimeout (Duration/ofSeconds 10)) .build))
worker
(when enabled
(future
(try
(while (:running? @state)
(let [generation (Object.)]
(swap! state assoc :generation generation :connected? false
:last-message (System/currentTimeMillis))
(try
(let [ws (connect! client uri query-fn state latest-price generation)]
(swap! state assoc :ws ws)
(try
(while (and (:running? @state) (:connected? @state)
(< (- (System/currentTimeMillis) (:last-message @state)) stale-after-ms))
(Thread/sleep 1000))
(finally (.abort ^WebSocket ws))))
(catch InterruptedException e (throw e))
(catch Exception e (log/warn e "Binance connection failed; retrying"))
(finally (swap! state assoc :connected? false :generation nil :ws nil))))
(when (:running? @state) (Thread/sleep 3000)))
(catch InterruptedException _))))]
{:state state :latest-price latest-price :worker worker}))
(defmethod ig/halt-key! :ws/binance
[_ {:keys [state]}]
(log/info "Stopping Binance WebSocket listener")
(swap! state assoc :running? false)
(when-let [ws (:ws @state)]
(try
(.sendClose ws WebSocket/NORMAL_CLOSURE "shutting down")
(catch Exception _))))
(defmethod ig/halt-key! :ws/binance [_ {:keys [state worker]}]
(swap! state assoc :running? false :connected? false :generation nil)
(when-let [ws (:ws @state)] (.abort ^WebSocket ws))
(when worker (future-cancel worker)))
@@ -1,26 +1,24 @@
(ns pmagnus.btcdata.web.controllers.health-test
(:require
[clojure.test :refer [deftest is testing]]
[pmagnus.btcdata.web.controllers.health :as health]))
(:require [clojure.test :refer [deftest is]]
[pmagnus.btcdata.web.controllers.health :as health]))
(def disabled-stream {:state (atom {:enabled? false}) :latest-price (atom nil)})
(deftest healthcheck-is-process-only
(let [response (health/healthcheck! {})]
(is (= 200 (:status response)))
(is (= "up" (get-in response [:body :app :status])))))
(is (= 200 (:status (health/healthcheck! {})))))
(deftest readiness-check-tests-database
(testing "ready when the database query succeeds"
(let [called (atom nil)
query-fn (fn [query params]
(reset! called [query params])
{:healthy 1})
response (health/readiness-check! query-fn {})]
(is (= 200 (:status response)))
(is (= {:status "ready"} (:body response)))
(is (= [:check-database {}] @called))))
(let [called (atom nil)
query-fn (fn [query params] (reset! called [query params]))]
(is (= 200 (:status (health/readiness-check! query-fn disabled-stream {}))))
(is (= [:check-database {}] @called)))
(is (= 503 (:status (health/readiness-check!
(fn [_ _] (throw (ex-info "database unavailable" {})))
disabled-stream {})))))
(testing "not ready when the database query fails"
(let [query-fn (fn [_ _] (throw (ex-info "database unavailable" {})))
response (health/readiness-check! query-fn {})]
(is (= 503 (:status response)))
(is (= {:status "not ready"} (:body response))))))
(deftest stale-stream-is-not-ready
(is (= 503 (:status (health/readiness-check!
(fn [_ _] {})
{:state (atom {:enabled? true :connected? true})
:latest-price (atom {:recorded-at java.time.Instant/EPOCH})}
{})))))
+52
View File
@@ -0,0 +1,52 @@
(ns pmagnus.btcdata.writer-test
(:require [clojure.test :refer [deftest is]]
[integrant.core :as ig]
[conman.core :as conman]
[next.jdbc :as jdbc]
[pmagnus.btcdata.writer :as writer]))
(deftest queries-use-lock-session
(let [connection (Object.)
seen (atom nil)
query (writer/guarded-query connection
(fn [& args] (reset! seen args) :result))]
(is (= :result (query :insert-price {:price 1})))
(is (= [connection :insert-price {:price 1}] @seen))))
(deftest postgres-session-fencing
;; Optional disposable PostgreSQL integration test; never use production here.
(if-let [url (System/getenv "BTCDATA_TEST_JDBC_URL")]
(let [ds (jdbc/get-datasource {:jdbcUrl url})
opts {:conn ds :query-fn (fn [conn _ _]
(jdbc/execute-one! conn ["select 1 as healthy"]))}
first-writer (ig/init-key :writer/query-fn opts)]
(try
(is (= 1 (:healthy (first-writer :check {}))))
(is (thrown-with-msg? clojure.lang.ExceptionInfo #"Another btcdata writer"
(ig/init-key :writer/query-fn opts)))
;; Closing the owning session cannot transparently reconnect for writes.
(.close (:connection (meta first-writer)))
(is (thrown? Exception (first-writer :check {})))
(let [replacement (ig/init-key :writer/query-fn opts)]
(try
(is (= 1 (:healthy (replacement :check {}))))
(finally (ig/halt-key! :writer/query-fn replacement))))
(finally
(when-not (.isClosed (:connection (meta first-writer)))
(ig/halt-key! :writer/query-fn first-writer)))))
(println "SKIP PostgreSQL fencing integration: set BTCDATA_TEST_JDBC_URL")))
(deftest pooled-session-releases-lock-on-halt
(when-let [url (System/getenv "BTCDATA_TEST_JDBC_URL")]
(let [pool (conman/connect! {:jdbc-url url :maximum-pool-size 3})
opts {:conn pool :query-fn (fn [_ _ _] :ok)}]
(try
(let [leader (ig/init-key :writer/query-fn opts)]
(is (thrown? Exception (ig/init-key :writer/query-fn opts)))
(ig/halt-key! :writer/query-fn leader))
;; An independent connection can acquire after pooled-session shutdown.
(let [other-opts (assoc opts :conn (jdbc/get-datasource {:jdbcUrl url}))
leader (ig/init-key :writer/query-fn other-opts)]
(is (= :ok (leader :check {})))
(ig/halt-key! :writer/query-fn leader))
(finally (conman/disconnect! pool))))))
@@ -0,0 +1,41 @@
(ns pmagnus.btcdata.ws.binance-test
(:require [clojure.test :refer [deftest is]]
[integrant.core :as ig]
[pmagnus.btcdata.ws.binance :as binance])
(:import [java.time Instant]))
(deftest freshness-boundaries
(let [state (atom {:enabled? true :connected? true})
price (atom {:recorded-at (Instant/ofEpochMilli 100000)})
component {:state state :latest-price price}]
(is (binance/fresh? component 100000))
(is (binance/fresh? component 160000))
(is (not (binance/fresh? component 160001)))
(is (not (binance/fresh? component 99999)))
(swap! state assoc :connected? false)
(is (not (binance/fresh? component 100000)))
(reset! price nil)
(is (not (binance/fresh? component 100000)))))
(deftest disabled-stream-does-not-connect
(let [component (ig/init-key :ws/binance {:enabled? "false" :uri "invalid"})]
(is (binance/fresh? component))
(is (nil? (:worker component)))
(ig/halt-key! :ws/binance component)
(is (false? (:running? @(:state component))))))
(deftest initial-failure-is-retried-and-worker-stops
(let [attempts (atom 0)
connect-var (ns-resolve 'pmagnus.btcdata.ws.binance 'connect!)]
(with-redefs-fn
{connect-var (fn [& _] (swap! attempts inc) (throw (ex-info "test failure" {})))}
(fn []
(let [component (ig/init-key :ws/binance {:enabled? "true" :uri "test"})]
(try
(let [deadline (+ (System/currentTimeMillis) 6000)]
(while (and (< @attempts 2) (< (System/currentTimeMillis) deadline))
(Thread/sleep 20)))
(is (>= @attempts 2))
(is (not (binance/fresh? component)))
(finally (ig/halt-key! :ws/binance component)))
(is (future-cancelled? (:worker component))))))))