This is an automated email from the ASF dual-hosted git repository.
morningman pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new 35fc172ba67 [Opt](docker) kerberos docker healthy check (#46662)
35fc172ba67 is described below
commit 35fc172ba676a68213f14a4ae03212da7fec5c87
Author: zgxme <zhenggaoxi...@selectdb.com>
AuthorDate: Fri Jan 10 18:01:04 2025 +0800
[Opt](docker) kerberos docker healthy check (#46662)
### What problem does this PR solve?
Previously, kerberos docker did not have health checks which could lead
to unstable tests.
---
.../kerberos/entrypoint-hive-master-2.sh | 20 +++++++++++++++++---
.../kerberos/entrypoint-hive-master.sh | 20 +++++++++++++++++---
.../kerberos/health-checks/hadoop-health-check.sh | 1 +
.../docker-compose/kerberos/health-checks/health.sh | 1 +
.../{health.sh => hive-health-check-2.sh} | 18 ++----------------
.../hive-health-check.sh} | 16 +---------------
.../docker-compose/kerberos/kerberos.yaml.tpl | 16 ++++++++++++----
docker/thirdparties/run-thirdparties-docker.sh | 14 +++++---------
8 files changed, 56 insertions(+), 50 deletions(-)
diff --git
a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master-2.sh
b/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master-2.sh
index c21460c3a57..eb95c5cb697 100755
--- a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master-2.sh
+++ b/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master-2.sh
@@ -25,12 +25,26 @@ cp /etc/trino/conf/presto-server.keytab
/keytabs/other-presto-server.keytab
cp /keytabs/update-location.sh /etc/hadoop-init.d/update-location.sh
/usr/local/hadoop-run.sh &
-sleep 30
+# check healthy hear
+echo "Waiting for hadoop to be healthy"
+
+for i in {1..10}; do
+ if /usr/local/health.sh; then
+ echo "Hadoop is healthy"
+ break
+ fi
+ echo "Hadoop is not healthy yet. Retrying in 20 seconds..."
+ sleep 20
+done
+
+if [ $i -eq 10 ]; then
+ echo "Hadoop did not become healthy after 120 attempts. Exiting."
+ exit 1
+fi
echo "Init kerberos test data"
kinit -kt /etc/hive/conf/hive.keytab hive/hadoop-maste...@otherrealm.com
hive -f /usr/local/sql/create_kerberos_hive_table.sql
-
-sleep 20
+touch /mnt/SUCCESS
tail -f /dev/null
diff --git
a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
b/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
index 62924992219..76f49724297 100755
--- a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
+++ b/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
@@ -23,12 +23,26 @@ mkdir -p /etc/hadoop-init.d/
cp /etc/trino/conf/* /keytabs/
/usr/local/hadoop-run.sh &
-sleep 30
+# check healthy hear
+echo "Waiting for hadoop to be healthy"
+
+for i in {1..10}; do
+ if /usr/local/health.sh; then
+ echo "Hadoop is healthy"
+ break
+ fi
+ echo "Hadoop is not healthy yet. Retrying in 20 seconds..."
+ sleep 20
+done
+
+if [ $i -eq 10 ]; then
+ echo "Hadoop did not become healthy after 120 attempts. Exiting."
+ exit 1
+fi
echo "Init kerberos test data"
kinit -kt /etc/hive/conf/hive.keytab hive/hadoop-mas...@labs.teradata.com
hive -f /usr/local/sql/create_kerberos_hive_table.sql
-
-sleep 20
+touch /mnt/SUCCESS
tail -f /dev/null
diff --git
a/docker/thirdparties/docker-compose/kerberos/health-checks/hadoop-health-check.sh
b/docker/thirdparties/docker-compose/kerberos/health-checks/hadoop-health-check.sh
index 190fa838d6f..77df431d85a 100755
---
a/docker/thirdparties/docker-compose/kerberos/health-checks/hadoop-health-check.sh
+++
b/docker/thirdparties/docker-compose/kerberos/health-checks/hadoop-health-check.sh
@@ -32,6 +32,7 @@ fi
FAILED=$(supervisorctl status | grep -v RUNNING || true)
if [ "$FAILED" == "" ]; then
+ echo "All services are running"
exit 0
else
echo "Some of the services are failing: ${FAILED}"
diff --git
a/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
b/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
old mode 100644
new mode 100755
index 515f37e36ac..473d7ceaeb6
--- a/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
+++ b/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
@@ -32,3 +32,4 @@ if test -d "${HEALTH_D}"; then
"${health_script}" &>> /var/log/container-health.log || exit 1
done
fi
+exit 0
diff --git
a/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
b/docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check-2.sh
old mode 100644
new mode 100755
similarity index 72%
copy from docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
copy to
docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check-2.sh
index 515f37e36ac..854524dac1f
--- a/docker/thirdparties/docker-compose/kerberos/health-checks/health.sh
+++
b/docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check-2.sh
@@ -16,19 +16,5 @@
# specific language governing permissions and limitations
# under the License.
-set -euo pipefail
-
-if test $# -gt 0; then
- echo "$0 does not accept arguments" >&2
- exit 32
-fi
-
-set -x
-
-HEALTH_D=${HEALTH_D:-/etc/health.d/}
-
-if test -d "${HEALTH_D}"; then
- for health_script in "${HEALTH_D}"/*; do
- "${health_script}" &>> /var/log/container-health.log || exit 1
- done
-fi
+kinit -kt /etc/hive/conf/hive.keytab hive/hadoop-maste...@otherrealm.com
+beeline -u
"jdbc:hive2://localhost:10000/default;principal=hive/hadoop-maste...@otherrealm.com"
-e "show databases;"
\ No newline at end of file
diff --git
a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
b/docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check.sh
similarity index 76%
copy from docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
copy to
docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check.sh
index 62924992219..4d3d86f69a2 100755
--- a/docker/thirdparties/docker-compose/kerberos/entrypoint-hive-master.sh
+++
b/docker/thirdparties/docker-compose/kerberos/health-checks/hive-health-check.sh
@@ -16,19 +16,5 @@
# specific language governing permissions and limitations
# under the License.
-set -euo pipefail
-
-echo "Copying kerberos keytabs to keytabs/"
-mkdir -p /etc/hadoop-init.d/
-cp /etc/trino/conf/* /keytabs/
-/usr/local/hadoop-run.sh &
-
-sleep 30
-
-echo "Init kerberos test data"
kinit -kt /etc/hive/conf/hive.keytab hive/hadoop-mas...@labs.teradata.com
-hive -f /usr/local/sql/create_kerberos_hive_table.sql
-
-sleep 20
-
-tail -f /dev/null
+beeline -u
"jdbc:hive2://localhost:10000/default;principal=hive/hadoop-mas...@labs.teradata.com"
-e "show databases;"
\ No newline at end of file
diff --git a/docker/thirdparties/docker-compose/kerberos/kerberos.yaml.tpl
b/docker/thirdparties/docker-compose/kerberos/kerberos.yaml.tpl
index 6aa353f3e0c..e635ed6bb27 100644
--- a/docker/thirdparties/docker-compose/kerberos/kerberos.yaml.tpl
+++ b/docker/thirdparties/docker-compose/kerberos/kerberos.yaml.tpl
@@ -24,13 +24,17 @@ services:
- ./sql:/usr/local/sql
-
./common/hadoop/apply-config-overrides.sh:/etc/hadoop-init.d/00-apply-config-overrides.sh
- ./common/hadoop/hadoop-run.sh:/usr/local/hadoop-run.sh
+ - ./health-checks/health.sh:/usr/local/health.sh
-
./health-checks/hadoop-health-check.sh:/etc/health.d/hadoop-health-check.sh
+ - ./health-checks/hive-health-check.sh:/etc/health.d/hive-health-check.sh
- ./entrypoint-hive-master.sh:/usr/local/entrypoint-hive-master.sh
- restart: on-failure
hostname: hadoop-master
entrypoint: /usr/local/entrypoint-hive-master.sh
healthcheck:
- test: ./health-checks/health.sh
+ test: ["CMD", "ls", "/mnt/SUCCESS"]
+ interval: 20s
+ timeout: 60s
+ retries: 120
ports:
- "5806:5006"
- "8820:8020"
@@ -46,17 +50,21 @@ services:
image: doristhirdpartydocker/trinodb:hdp3.1-hive-kerberized-2_96
container_name: doris--kerberos2
hostname: hadoop-master-2
- restart: on-failure
volumes:
- ./two-kerberos-hives:/keytabs
- ./sql:/usr/local/sql
-
./common/hadoop/apply-config-overrides.sh:/etc/hadoop-init.d/00-apply-config-overrides.sh
- ./common/hadoop/hadoop-run.sh:/usr/local/hadoop-run.sh
+ - ./health-checks/health.sh:/usr/local/health.sh
-
./health-checks/hadoop-health-check.sh:/etc/health.d/hadoop-health-check.sh
+ -
./health-checks/hive-health-check-2.sh:/etc/health.d/hive-health-check-2.sh
- ./entrypoint-hive-master-2.sh:/usr/local/entrypoint-hive-master-2.sh
entrypoint: /usr/local/entrypoint-hive-master-2.sh
healthcheck:
- test: ./health-checks/health.sh
+ test: ["CMD", "ls", "/mnt/SUCCESS"]
+ interval: 20s
+ timeout: 60s
+ retries: 120
ports:
- "15806:5006"
- "18820:8020"
diff --git a/docker/thirdparties/run-thirdparties-docker.sh
b/docker/thirdparties/run-thirdparties-docker.sh
index fd9558eef55..e3769025fec 100755
--- a/docker/thirdparties/run-thirdparties-docker.sh
+++ b/docker/thirdparties/run-thirdparties-docker.sh
@@ -708,6 +708,11 @@ if [[ "${RUN_MINIO}" -eq 1 ]]; then
pids["minio"]=$!
fi
+if [[ "${RUN_KERBEROS}" -eq 1 ]]; then
+ start_kerberos > start_kerberos.log 2>&1 &
+ pids["kerberos"]=$!
+fi
+
echo "waiting all dockers starting done"
for compose in "${!pids[@]}"; do
@@ -727,15 +732,6 @@ for compose in "${!pids[@]}"; do
fi
done
-if [[ "${RUN_KERBEROS}" -eq 1 ]]; then
- echo "Starting Kerberos after all other components..."
- start_kerberos > start_kerberos.log 2>&1
- if [ $? -ne 0 ]; then
- echo "Kerberos startup failed"
- cat start_kerberos.log
- exit 1
- fi
-fi
echo "docker started"
docker ps -a --format "{{.ID}} | {{.Image}} | {{.Status}}"
echo "all dockers started successfully"
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
