Details
-
Bug
-
Status: Stalled (View Workflow)
-
Major
-
Resolution: Unresolved
-
13.1, 10.11.19, 11.4.13, 11.8.9, 12.3.3
-
Linux x86_64 host; official image `mariadb@sha256:dd9b303aed4f4890ed09f766d8ca9ddfd176c0c6f6267feff53b3192ec65a979`; server reports `12.3.3-MariaDB-ubu2404`, source revision `83e909fc2a0dbc394b4b683fb3fa2d7dcf26cc5e`.
Description
When an InnoDB encryption worker has pending default-encryption work and
`innodb_encryption_rotation_iops` is zero, setting
`innodb_encryption_threads=0` does not finish. The client remains in the
statement while the server continues to answer other connections.
Steps to reproduce
IMAGE='mariadb@sha256:dd9b303aed4f4890ed09f766d8ca9ddfd176c0c6f6267feff53b3192ec65a979'
|
C='mdev-encryption-worker-repro'
|
D=$(mktemp -d)
|
cleanup() {
|
docker rm -f "$C" >/dev/null 2>&1 || true
|
rm -r "$D" >/dev/null 2>&1 || true
|
}
|
trap cleanup EXIT
|
printf '%s\n' '1;11111111111111111111111111111111' > "$D/keys.txt"
|
chmod 0644 "$D/keys.txt"
|
|
|
docker run --detach --name "$C" \
|
--mount type=bind,src="$D/keys.txt",dst=/keys.txt,readonly \
|
--env MARIADB_ALLOW_EMPTY_ROOT_PASSWORD=1 \
|
"$IMAGE" \
|
--plugin-load-add=file_key_management.so \
|
--file-key-management-filename=/keys.txt
|
|
|
for i in $(seq 1 90); do
|
if docker exec "$C" mariadb --protocol=socket -uroot -NBe \
|
"SELECT @@port" 2>/dev/null | grep -qx 3306; then
|
break
|
fi
|
sleep 1
|
done
|
|
|
docker exec "$C" mariadb --protocol=socket -uroot -e "
|
SELECT PLUGIN_STATUS
|
FROM INFORMATION_SCHEMA.PLUGINS
|
WHERE PLUGIN_NAME='file_key_management';
|
SET GLOBAL innodb_encrypt_tables='OFF';
|
CREATE DATABASE crypt_scope;
|
CREATE TABLE crypt_scope.t4(
|
id INT PRIMARY KEY,
|
pad VARCHAR(4096)
|
) ENGINE=InnoDB ENCRYPTED=NO;
|
INSERT INTO crypt_scope.t4
|
SELECT seq, REPEAT('x',4000) FROM seq_1_to_10000;
|
FLUSH TABLES crypt_scope.t4;
|
SET GLOBAL innodb_encryption_threads=1;
|
SET GLOBAL innodb_encryption_rotation_iops=0;
|
SET GLOBAL innodb_encrypt_tables='ON';
|
DO SLEEP(2);
|
"
|
|
|
# This statement should return promptly, but it remains blocked.
|
timeout 8 docker exec "$C" mariadb --protocol=socket -uroot -e \
|
"SET GLOBAL innodb_encryption_threads=0"
|
echo "shutdown_statement_rc=$?"
|
|
|
sleep 20
|
docker exec "$C" mariadb --protocol=socket -uroot -e "SHOW FULL PROCESSLIST"
|
docker exec "$C" mariadb-admin --protocol=socket -uroot ping
|
|
For comparison, repeat the test with
`innodb_encryption_rotation_iops=100`; the final `SET GLOBAL` returns normally.
Actual result
In the reproduced 12.3.3 run, the zero-IOPS case exceeded the 8-second bound:
shutdown_statement_rc=124
|
`SHOW FULL PROCESSLIST` still showed
`SET GLOBAL innodb_encryption_threads=0` after 29 seconds. At the same time:
mysqld is alive
|
The `rotation_iops=100` control completed within the same 8-second bound.
Expected result
Setting `innodb_encryption_threads=0` should stop the worker and return in
bounded time even when the configured rotation IOPS value is zero. It should
not require another session to change the IOPS value or a server restart.
Impact and boundary
An administrator can leave the configuration statement and encryption worker
shutdown indefinitely blocked. The server process itself did not crash and
continued to answer health checks. Only MariaDB 12.3.3 was tested. This report
is researcher-reproduced and not yet maintainer-confirmed.
Attachments
Issue Links
- is caused by
-
-
- Closed
-
