From eea0e09b6dae634ff95f0862aeff45a1a45c7d2f Mon Sep 17 00:00:00 2001
From: Valery Kharseko <vharseko@3a-systems.ru>
Date: Mon, 28 Sep 2026 12:49:14 +0000
Subject: [PATCH] [#1092] Probe the Docker container's health without binding as the root user (#1102)
---
.github/workflows/build.yml | 136 ++++++++++++++++++++++++++++++++++++++++----
1 files changed, 122 insertions(+), 14 deletions(-)
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 4c54e32..ea69ca8 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -586,10 +586,10 @@
cleanup() { docker rm -f test_master $REPLICAS >/dev/null 2>&1 || true; docker network rm test_replication >/dev/null 2>&1 || true; }
cleanup
trap 'code=$?; for c in test_master $REPLICAS; do echo "::group::container logs ($c)"; docker logs $c 2>&1 || true; echo "::endgroup::"; done; cleanup; exit $code' ERR
- # every tool reads the root password from a file (#1084); dsreplication run with -n prints
+ # every tool reads the root password from a file (#1084, #1092); dsreplication run with -n prints
# no command line, so a password put back on one would pass every check below
- rc=0; docker run --rm --entrypoint grep "$IMAGE" -nE -- '(^|[[:space:]])(-w|--(bindPassword[12]?|adminPassword))([[:space:]=]|$)' /opt/opendj/bootstrap/replicate.sh || rc=$?
- if [ $rc -ne 1 ]; then echo "::error::replicate.sh passes the root password on a command line, or grep could not read it"; false; fi
+ rc=0; docker run --rm --entrypoint grep "$IMAGE" -nE -- '(^|[[:space:]])(-w|--(bindPassword[12]?|adminPassword|rootUserPassword))([[:space:]=]|$)' /opt/opendj/bootstrap/setup.sh /opt/opendj/bootstrap/replicate.sh || rc=$?
+ if [ $rc -ne 1 ]; then echo "::error::setup.sh or replicate.sh passes the root password on a command line, or grep could not read them"; false; fi
# the password file goes to /dev/shm, off the writable layer of the container, and the mktemp of the image puts it there
docker run --rm --entrypoint grep "$IMAGE" -qF -- 'mktemp -p /dev/shm "opendj-replicate.$ADMIN_PORT.' /opt/opendj/bootstrap/replicate.sh || { echo "::error::replicate.sh no longer puts the password file on /dev/shm"; false; }
docker run --rm --entrypoint sh "$IMAGE" -c 'f=$(mktemp -p /dev/shm "opendj-replicate.$ADMIN_PORT.XXXXXX") && rm -f "$f" && case $f in /dev/shm/opendj-replicate.4444.*) ;; *) exit 1;; esac' || { echo "::error::mktemp in the image does not create the password file on /dev/shm"; false; }
@@ -636,13 +636,14 @@
if [ $rc -ne 5 ] || grep -qE "trying again|initializing replication" <<<"$out"; then
echo "$out"; echo "::error::a second replicate.sh exited with $rc, not with the 5 of its dsreplication enable, or went on after it"; false
fi
- # the root password shows in no container log, and the file replicate.sh passed it in is gone (#1084)
+ # the root password shows in no container log, and the files setup.sh and replicate.sh passed it in are gone (#1084, #1092)
for c in test_master $REPLICAS; do
if docker logs $c 2>&1 | grep -F "$ROOT_PASSWORD"; then echo "::error::The root password is in the log of $c"; false; fi
done
- for c in $REPLICAS; do
- # the JVM of the HEALTHCHECK's ldapsearch keeps its command line, root password included, in /tmp/hsperfdata_* while it runs
- left=$(docker exec $c grep -rlsF -- "$ROOT_PASSWORD" /tmp /dev/shm | grep -v '^/tmp/hsperfdata_' || true)
+ for c in test_master $REPLICAS; do
+ # a JVM keeps its command line in /tmp/hsperfdata_* while it runs; the HEALTHCHECK no longer binds as root (#1092),
+ # so no process left running has the root password on it
+ left=$(docker exec $c grep -rlsF -- "$ROOT_PASSWORD" /tmp /dev/shm || true)
if [ -n "$left" ]; then echo "::error::The root password is left in $left of $c"; false; fi
done
docker exec test_replica test -e /dev/shm/opendj-replicate.5444.other || { echo "::error::run.sh of test_replica removed the password file of another container"; false; }
@@ -705,6 +706,59 @@
done
docker exec test_bootstrap test -e "$shm_other" || { echo "::error::run.sh removed $shm_other, the password file of another container"; false; }
docker kill test_bootstrap test_bootstrap_shm
+ - name: Docker test health check
+ shell: bash
+ run: |
+ # the ERR trap below has to fire for a check failing inside stays_healthy too
+ set -o errtrace
+ # the containers are run without --rm, so that one whose bootstrap failed is still there for the trap to print
+ trap 'code=$?; for c in test_health test_health_bind; do echo "::group::container logs ($c)"; docker logs $c 2>&1 || true; docker inspect --format="{{json .State.Health}}" $c 2>&1 || true; echo "::endgroup::"; done; docker rm -f test_health test_health_bind >/dev/null 2>&1 || true; exit $code' ERR
+ IMAGE=localhost:5000/${GITHUB_REPOSITORY,,}:${{ env.release_version }}
+ # a failed probe leaves a failing streak until the next probe passes, 5 s later at the
+ # earliest, so a container found "healthy 0" every 2 s for 45 s passed every probe since
+ stays_healthy() {
+ local end=$((SECONDS + 45))
+ while [ $SECONDS -lt $end ]; do
+ test "$(docker inspect --format='{{.State.Health.Status}} {{.State.Health.FailingStreak}}' "$1")" = "healthy 0"
+ sleep 2
+ done
+ }
+ # ROOT_PASSWORD is only the initial root password: changing it must not turn the container unhealthy
+ docker run -it -d --memory="512m" --health-interval=5s -e ROOT_PASSWORD=initial_password --name=test_health $IMAGE
+ timeout 3m bash -c 'until docker inspect --format="{{json .State.Health.Status}}" test_health | grep -q \"healthy\"; do sleep 10; done'
+ docker exec test_health /opt/opendj/bin/ldappasswordmodify --hostname localhost --port 1636 --useSsl --trustAll --bindDN "cn=Directory Manager" --bindPassword initial_password --currentPassword initial_password --newPassword rotated_password
+ # bind settings kept for the CLI in the home of the image user must not reach the probe
+ docker exec test_health sh -c 'mkdir -p /home/opendj/.opendj && printf "bindDN=cn=Directory Manager\nbindPassword=wrong_password\n" > /home/opendj/.opendj/tools.properties'
+ stays_healthy test_health
+ docker rm -f test_health
+ # an instance rejecting unauthenticated requests is probed with the account it is given, whose password is read from a file
+ # a password of its own, so the command lines below can be searched for it
+ printf hc_secret_1092 > "$RUNNER_TEMP/healthcheck_password"
+ chmod 644 "$RUNNER_TEMP/healthcheck_password"
+ docker run -it -d --memory="512m" --health-interval=5s -v "$RUNNER_TEMP/healthcheck_password:/tmp/healthcheck_password:ro" -e ROOT_PASSWORD=hc_secret_1092 -e HEALTHCHECK_BIND_DN="cn=Directory Manager" -e HEALTHCHECK_BIND_PASSWORD_FILE=/tmp/healthcheck_password --name=test_health_bind $IMAGE
+ timeout 3m bash -c 'until docker inspect --format="{{json .State.Health.Status}}" test_health_bind | grep -q \"healthy\"; do sleep 10; done'
+ docker exec test_health_bind /opt/opendj/bin/dsconfig set-global-configuration-prop --hostname localhost --port 4444 --bindDN "cn=Directory Manager" --bindPasswordFile /tmp/healthcheck_password --set reject-unauthenticated-requests:true --no-prompt --trustAll
+ # the setting has taken: the anonymous probe would now be refused
+ rc=0
+ docker exec test_health_bind /opt/opendj/bin/ldapsearch --hostname localhost --port 1636 --useSsl --trustAll --baseDN "" --searchScope base "(objectClass=*)" 1.1 || rc=$?
+ test "$rc" = 53
+ # and the probe reports that refusal when it has no account to bind with
+ rc=0
+ docker exec -e HEALTHCHECK_BIND_DN= test_health_bind /opt/opendj/healthcheck.sh || rc=$?
+ test "$rc" = 1
+ # a password file it cannot read is reported as such, whether it is missing or there but not
+ # readable: the image runs as its own user, not root, so mode 000 keeps the probe out
+ docker exec test_health_bind sh -c 'touch /tmp/unreadable_password && chmod 000 /tmp/unreadable_password'
+ for f in /nonexistent /tmp/unreadable_password; do
+ rc=0
+ out=$(docker exec -e HEALTHCHECK_BIND_PASSWORD_FILE=$f test_health_bind /opt/opendj/healthcheck.sh) || rc=$?
+ test "$rc" = 1
+ grep -q 'is not a readable file' <<< "$out"
+ done
+ stays_healthy test_health_bind
+ # the password never shows on a command line: sample every process's for two probe intervals
+ docker exec test_health_bind sh -c 'end=$(($(date +%s) + 12)); while [ "$(date +%s)" -lt "$end" ]; do for f in /proc/[0-9]*/cmdline; do tr "\0" " " < "$f" 2>/dev/null; echo; done | grep -q "[h]c_secret_1092" && exit 1; sleep 0.2; done; exit 0'
+ docker rm -f test_health_bind
- name: Scan image for vulnerabilities (Trivy)
# trivy resolves the image from the local Docker daemon, so only the runner's
# linux/amd64 manifest is scanned; cache: false keeps the ~1GB trivy DBs from
@@ -894,10 +948,10 @@
cleanup() { docker rm -f test_master $REPLICAS >/dev/null 2>&1 || true; docker network rm test_replication >/dev/null 2>&1 || true; }
cleanup
trap 'code=$?; for c in test_master $REPLICAS; do echo "::group::container logs ($c)"; docker logs $c 2>&1 || true; echo "::endgroup::"; done; cleanup; exit $code' ERR
- # every tool reads the root password from a file (#1084); dsreplication run with -n prints
+ # every tool reads the root password from a file (#1084, #1092); dsreplication run with -n prints
# no command line, so a password put back on one would pass every check below
- rc=0; docker run --rm --entrypoint grep "$IMAGE" -nE -- '(^|[[:space:]])(-w|--(bindPassword[12]?|adminPassword))([[:space:]=]|$)' /opt/opendj/bootstrap/replicate.sh || rc=$?
- if [ $rc -ne 1 ]; then echo "::error::replicate.sh passes the root password on a command line, or grep could not read it"; false; fi
+ rc=0; docker run --rm --entrypoint grep "$IMAGE" -nE -- '(^|[[:space:]])(-w|--(bindPassword[12]?|adminPassword|rootUserPassword))([[:space:]=]|$)' /opt/opendj/bootstrap/setup.sh /opt/opendj/bootstrap/replicate.sh || rc=$?
+ if [ $rc -ne 1 ]; then echo "::error::setup.sh or replicate.sh passes the root password on a command line, or grep could not read them"; false; fi
# the password file goes to /dev/shm, off the writable layer of the container, and the mktemp of the image puts it there
docker run --rm --entrypoint grep "$IMAGE" -qF -- 'mktemp -p /dev/shm "opendj-replicate.$ADMIN_PORT.' /opt/opendj/bootstrap/replicate.sh || { echo "::error::replicate.sh no longer puts the password file on /dev/shm"; false; }
docker run --rm --entrypoint sh "$IMAGE" -c 'f=$(mktemp -p /dev/shm "opendj-replicate.$ADMIN_PORT.XXXXXX") && rm -f "$f" && case $f in /dev/shm/opendj-replicate.4444.*) ;; *) exit 1;; esac' || { echo "::error::mktemp in the image does not create the password file on /dev/shm"; false; }
@@ -944,13 +998,14 @@
if [ $rc -ne 5 ] || grep -qE "trying again|initializing replication" <<<"$out"; then
echo "$out"; echo "::error::a second replicate.sh exited with $rc, not with the 5 of its dsreplication enable, or went on after it"; false
fi
- # the root password shows in no container log, and the file replicate.sh passed it in is gone (#1084)
+ # the root password shows in no container log, and the files setup.sh and replicate.sh passed it in are gone (#1084, #1092)
for c in test_master $REPLICAS; do
if docker logs $c 2>&1 | grep -F "$ROOT_PASSWORD"; then echo "::error::The root password is in the log of $c"; false; fi
done
- for c in $REPLICAS; do
- # the JVM of the HEALTHCHECK's ldapsearch keeps its command line, root password included, in /tmp/hsperfdata_* while it runs
- left=$(docker exec $c grep -rlsF -- "$ROOT_PASSWORD" /tmp /dev/shm | grep -v '^/tmp/hsperfdata_' || true)
+ for c in test_master $REPLICAS; do
+ # a JVM keeps its command line in /tmp/hsperfdata_* while it runs; the HEALTHCHECK no longer binds as root (#1092),
+ # so no process left running has the root password on it
+ left=$(docker exec $c grep -rlsF -- "$ROOT_PASSWORD" /tmp /dev/shm || true)
if [ -n "$left" ]; then echo "::error::The root password is left in $left of $c"; false; fi
done
docker exec test_replica test -e /dev/shm/opendj-replicate.5444.other || { echo "::error::run.sh of test_replica removed the password file of another container"; false; }
@@ -1013,6 +1068,59 @@
done
docker exec test_bootstrap test -e "$shm_other" || { echo "::error::run.sh removed $shm_other, the password file of another container"; false; }
docker kill test_bootstrap test_bootstrap_shm
+ - name: Docker test health check
+ shell: bash
+ run: |
+ # the ERR trap below has to fire for a check failing inside stays_healthy too
+ set -o errtrace
+ # the containers are run without --rm, so that one whose bootstrap failed is still there for the trap to print
+ trap 'code=$?; for c in test_health test_health_bind; do echo "::group::container logs ($c)"; docker logs $c 2>&1 || true; docker inspect --format="{{json .State.Health}}" $c 2>&1 || true; echo "::endgroup::"; done; docker rm -f test_health test_health_bind >/dev/null 2>&1 || true; exit $code' ERR
+ IMAGE=localhost:5000/${GITHUB_REPOSITORY,,}:${{ env.release_version }}-alpine
+ # a failed probe leaves a failing streak until the next probe passes, 5 s later at the
+ # earliest, so a container found "healthy 0" every 2 s for 45 s passed every probe since
+ stays_healthy() {
+ local end=$((SECONDS + 45))
+ while [ $SECONDS -lt $end ]; do
+ test "$(docker inspect --format='{{.State.Health.Status}} {{.State.Health.FailingStreak}}' "$1")" = "healthy 0"
+ sleep 2
+ done
+ }
+ # ROOT_PASSWORD is only the initial root password: changing it must not turn the container unhealthy
+ docker run -it -d --memory="1g" --health-interval=5s -e ROOT_PASSWORD=initial_password --name=test_health $IMAGE
+ timeout 3m bash -c 'until docker inspect --format="{{json .State.Health.Status}}" test_health | grep -q \"healthy\"; do sleep 10; done'
+ docker exec test_health /opt/opendj/bin/ldappasswordmodify --hostname localhost --port 1636 --useSsl --trustAll --bindDN "cn=Directory Manager" --bindPassword initial_password --currentPassword initial_password --newPassword rotated_password
+ # bind settings kept for the CLI in the home of the image user must not reach the probe
+ docker exec test_health sh -c 'mkdir -p /home/opendj/.opendj && printf "bindDN=cn=Directory Manager\nbindPassword=wrong_password\n" > /home/opendj/.opendj/tools.properties'
+ stays_healthy test_health
+ docker rm -f test_health
+ # an instance rejecting unauthenticated requests is probed with the account it is given, whose password is read from a file
+ # a password of its own, so the command lines below can be searched for it
+ printf hc_secret_1092 > "$RUNNER_TEMP/healthcheck_password"
+ chmod 644 "$RUNNER_TEMP/healthcheck_password"
+ docker run -it -d --memory="1g" --health-interval=5s -v "$RUNNER_TEMP/healthcheck_password:/tmp/healthcheck_password:ro" -e ROOT_PASSWORD=hc_secret_1092 -e HEALTHCHECK_BIND_DN="cn=Directory Manager" -e HEALTHCHECK_BIND_PASSWORD_FILE=/tmp/healthcheck_password --name=test_health_bind $IMAGE
+ timeout 3m bash -c 'until docker inspect --format="{{json .State.Health.Status}}" test_health_bind | grep -q \"healthy\"; do sleep 10; done'
+ docker exec test_health_bind /opt/opendj/bin/dsconfig set-global-configuration-prop --hostname localhost --port 4444 --bindDN "cn=Directory Manager" --bindPasswordFile /tmp/healthcheck_password --set reject-unauthenticated-requests:true --no-prompt --trustAll
+ # the setting has taken: the anonymous probe would now be refused
+ rc=0
+ docker exec test_health_bind /opt/opendj/bin/ldapsearch --hostname localhost --port 1636 --useSsl --trustAll --baseDN "" --searchScope base "(objectClass=*)" 1.1 || rc=$?
+ test "$rc" = 53
+ # and the probe reports that refusal when it has no account to bind with
+ rc=0
+ docker exec -e HEALTHCHECK_BIND_DN= test_health_bind /opt/opendj/healthcheck.sh || rc=$?
+ test "$rc" = 1
+ # a password file it cannot read is reported as such, whether it is missing or there but not
+ # readable: the image runs as its own user, not root, so mode 000 keeps the probe out
+ docker exec test_health_bind sh -c 'touch /tmp/unreadable_password && chmod 000 /tmp/unreadable_password'
+ for f in /nonexistent /tmp/unreadable_password; do
+ rc=0
+ out=$(docker exec -e HEALTHCHECK_BIND_PASSWORD_FILE=$f test_health_bind /opt/opendj/healthcheck.sh) || rc=$?
+ test "$rc" = 1
+ grep -q 'is not a readable file' <<< "$out"
+ done
+ stays_healthy test_health_bind
+ # the password never shows on a command line: sample every process's for two probe intervals
+ docker exec test_health_bind sh -c 'end=$(($(date +%s) + 12)); while [ "$(date +%s)" -lt "$end" ]; do for f in /proc/[0-9]*/cmdline; do tr "\0" " " < "$f" 2>/dev/null; echo; done | grep -q "[h]c_secret_1092" && exit 1; sleep 0.2; done; exit 0'
+ docker rm -f test_health_bind
- name: Scan image for vulnerabilities (Trivy)
# trivy resolves the image from the local Docker daemon, so only the runner's
# linux/amd64 manifest is scanned; cache: false keeps the ~1GB trivy DBs from
--
Gitblit v1.10.0