skip inconsistent shards if possible (#13945)

data shards were wrong due to a healing bug
reported in #13803 mainly with unaligned object
sizes.

This PR is an attempt to automatically avoid
these shards, with available information about
the `xl.meta` and actually disk mtime.
This commit is contained in:
Harshavardhana
2021-12-21 10:08:26 -08:00
committed by GitHub
parent 6fbf4f96b6
commit 0e3037631f
17 changed files with 548 additions and 129 deletions

View File

@@ -21,51 +21,70 @@ function start_minio_3_node() {
start_port=$2
args=""
for i in $(seq 1 3); do
args="$args http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/1/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/2/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/3/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/4/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/5/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/6/"
args="$args http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/1/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/2/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/3/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/4/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/5/ http://127.0.0.1:$[$start_port+$i]${WORK_DIR}/$i/6/"
done
"${MINIO[@]}" --address ":$[$start_port+1]" $args > "${WORK_DIR}/dist-minio-server1.log" 2>&1 &
disown $!
pid1=$!
disown ${pid1}
"${MINIO[@]}" --address ":$[$start_port+2]" $args > "${WORK_DIR}/dist-minio-server2.log" 2>&1 &
disown $!
pid2=$!
disown $pid2
"${MINIO[@]}" --address ":$[$start_port+3]" $args > "${WORK_DIR}/dist-minio-server3.log" 2>&1 &
disown $!
pid3=$!
disown $pid3
sleep "$1"
if [ "$(pgrep -c minio)" -ne 3 ]; then
for i in $(seq 1 3); do
echo "server$i log:"
cat "${WORK_DIR}/dist-minio-server$i.log"
done
echo "FAILED"
purge "$WORK_DIR"
exit 1
if ! ps -p $pid1 1>&2 > /dev/null; then
echo "server1 log:"
cat "${WORK_DIR}/dist-minio-server1.log"
echo "FAILED"
purge "$WORK_DIR"
exit 1
fi
if ! ps -p $pid2 1>&2 > /dev/null; then
echo "server2 log:"
cat "${WORK_DIR}/dist-minio-server2.log"
echo "FAILED"
purge "$WORK_DIR"
exit 1
fi
if ! ps -p $pid3 1>&2 > /dev/null; then
echo "server3 log:"
cat "${WORK_DIR}/dist-minio-server3.log"
echo "FAILED"
purge "$WORK_DIR"
exit 1
fi
if ! pkill minio; then
for i in $(seq 1 3); do
echo "server$i log:"
cat "${WORK_DIR}/dist-minio-server$i.log"
done
echo "FAILED"
purge "$WORK_DIR"
exit 1
for i in $(seq 1 3); do
echo "server$i log:"
cat "${WORK_DIR}/dist-minio-server$i.log"
done
echo "FAILED"
purge "$WORK_DIR"
exit 1
fi
sleep 1;
if pgrep minio; then
# forcibly killing, to proceed further properly.
if ! pkill -9 minio; then
echo "no minio process running anymore, proceed."
fi
# forcibly killing, to proceed further properly.
if ! pkill -9 minio; then
echo "no minio process running anymore, proceed."
fi
fi
}
function check_online() {
if grep -q 'Unable to initialize sub-systems' ${WORK_DIR}/dist-minio-*.log; then
echo "1"
echo "1"
fi
}
@@ -96,14 +115,14 @@ function perform_test() {
rv=$(check_online)
if [ "$rv" == "1" ]; then
for i in $(seq 1 3); do
echo "server$i log:"
cat "${WORK_DIR}/dist-minio-server$i.log"
done
pkill -9 minio
echo "FAILED"
purge "$WORK_DIR"
exit 1
for i in $(seq 1 3); do
echo "server$i log:"
cat "${WORK_DIR}/dist-minio-server$i.log"
done
pkill -9 minio
echo "FAILED"
purge "$WORK_DIR"
exit 1
fi
}