From 60242f30b8d1c07cfd764eb7b0eb48db9d20ed34 Mon Sep 17 00:00:00 2001
From: JSONbored <49853598+JSONbored@users.noreply.github.com>
Date: Sat, 11 Jul 2026 06:58:45 -0700
Subject: [PATCH] fix(selfhost): retry the /ready probe in
 selfhost-post-update-check.sh instead of a single attempt (#5085)

docker compose up -d returns as soon as the container starts, well
before the app inside has finished booting and bound its port, so a
single immediate curl reliably false-failed on a completely normal
deploy -- hit twice today. The gittensory service's own Docker
healthcheck already documents and tolerates this (start_period: 60s,
"tolerates the Postgres cold start"); the post-update script had no
equivalent tolerance. Retries up to 90s by default (configurable via
SELFHOST_READY_RETRIES/SELFHOST_READY_RETRY_DELAY_SECONDS), verified
against a local mock server that fails twice before succeeding, and
against one that never succeeds (exhausts and exits non-zero).
---
 scripts/selfhost-post-update-check.sh | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git a/scripts/selfhost-post-update-check.sh b/scripts/selfhost-post-update-check.sh
index 42759d8d35..cd85d386cf 100755
--- a/scripts/selfhost-post-update-check.sh
+++ b/scripts/selfhost-post-update-check.sh
@@ -30,9 +30,24 @@ if [ -z "$container_id" ]; then
   exit 1
 fi
 
+# Retry, don't single-probe: `docker compose up -d` returns as soon as the container STARTS, well before
+# the app inside has bound its port -- a single immediate curl reliably false-fails on a normal boot. Budget
+# matches the gittensory service's own Docker healthcheck start_period (60s, docker-compose.yml) plus margin,
+# polling often enough that a normal ~15-20s boot returns almost immediately once actually ready.
+READY_RETRIES="${SELFHOST_READY_RETRIES:-45}"
+READY_RETRY_DELAY_SECONDS="${SELFHOST_READY_RETRY_DELAY_SECONDS:-2}"
+
 echo "selfhost post-update check: probing $READY_URL"
-if ! curl -sf "$READY_URL" >/dev/null; then
-  echo "error: $READY_URL did not return HTTP 2xx" >&2
+ready=0
+for _ in $(seq 1 "$READY_RETRIES"); do
+  if curl -sf "$READY_URL" >/dev/null; then
+    ready=1
+    break
+  fi
+  sleep "$READY_RETRY_DELAY_SECONDS"
+done
+if [ "$ready" -ne 1 ]; then
+  echo "error: $READY_URL did not return HTTP 2xx after $READY_RETRIES attempts ($((READY_RETRIES * READY_RETRY_DELAY_SECONDS))s)" >&2
   exit 1
 fi