thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-12-18 04:12:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #23161 from ClickHouse/logs_for_debuging_test_failures

Browse Source 
More verbose logs for debuging test failures with Replicated and Keeper
This commit is contained in:
tavplubix 2021-04-16 12:01:58 +03:00 committed by GitHub
commit ae3dc3dfc0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 29 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
src/Interpreters/DDLWorker.cpp
View File

@ -372,7 +372,20 @@ void DDLWorker::scheduleTasks(bool reinitialized)
}
Strings queue_nodes = zookeeper->getChildren(queue_dir, nullptr, queue_updated_event);
size_t size_before_filtering = queue_nodes.size();
filterAndSortQueueNodes(queue_nodes);
/// The following message is too verbose, but it can be useful too debug mysterious test failures in CI
LOG_TRACE(log, "scheduleTasks: initialized={}, size_before_filtering={}, queue_size={}, "
"entries={}..{}, "
"first_failed_task_name={}, current_tasks_size={},"
"last_current_task={},"
"last_skipped_entry_name={}",
initialized, size_before_filtering, queue_nodes.size(),
queue_nodes.empty() ? "none" : queue_nodes.front(), queue_nodes.empty() ? "none" : queue_nodes.back(),
first_failed_task_name ? *first_failed_task_name : "none", current_tasks.size(),
current_tasks.empty() ? "none" : current_tasks.back()->entry_name,
last_skipped_entry_name ? *last_skipped_entry_name : "none");
if (max_tasks_in_queue < queue_nodes.size())
cleanup_event->set();

13
src/Storages/MergeTree/ReplicatedMergeTreeCleanupThread.cpp
View File

@ -342,6 +342,15 @@ void ReplicatedMergeTreeCleanupThread::clearOldBlocks()
timed_blocks.begin(), timed_blocks.end(), block_threshold, NodeWithStat::greaterByTime);
auto first_outdated_block = std::min(first_outdated_block_fixed_threshold, first_outdated_block_time_threshold);
auto num_nodes_to_delete = timed_blocks.end() - first_outdated_block;
if (!num_nodes_to_delete)
return;
auto last_outdated_block = timed_blocks.end() - 1;
LOG_TRACE(log, "Will clear {} old blocks from {} (ctime {}) to {} (ctime {})", num_nodes_to_delete,
first_outdated_block->node, first_outdated_block->ctime,
last_outdated_block->node, last_outdated_block->ctime);
zkutil::AsyncResponses<Coordination::RemoveResponse> try_remove_futures;
for (auto it = first_outdated_block; it != timed_blocks.end(); ++it)
{
@ -372,9 +381,7 @@ void ReplicatedMergeTreeCleanupThread::clearOldBlocks()
first_outdated_block++;
}
auto num_nodes_to_delete = timed_blocks.end() - first_outdated_block;
if (num_nodes_to_delete)
LOG_TRACE(log, "Cleared {} old blocks from ZooKeeper", num_nodes_to_delete);
LOG_TRACE(log, "Cleared {} old blocks from ZooKeeper", num_nodes_to_delete);
}

8
tests/queries/0_stateless/00953_zookeeper_suetin_deduplication_bug.sh
View File

@ -21,15 +21,12 @@ ORDER BY (engine_id)
SETTINGS replicated_deduplication_window = 2, cleanup_delay_period=4, cleanup_delay_period_random_add=0;"
$CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 1, 'hello')"
sleep 1
$CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 2, 'hello')"
sleep 1
$CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 3, 'hello')"
$CLICKHOUSE_CLIENT --query="SELECT count(*) from elog" # 3 rows
count=$($CLICKHOUSE_CLIENT --query="SELECT COUNT(*) FROM system.zookeeper where path = '/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/elog/s1/blocks'")
while [[ $count != 2 ]]
do
sleep 1
@ -39,9 +36,8 @@ done
$CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 1, 'hello')"
$CLICKHOUSE_CLIENT --query="SELECT count(*) from elog" # 4 rows
count=$($CLICKHOUSE_CLIENT --query="SELECT COUNT(*) FROM system.zookeeper where path = '/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/elog/s1/blocks'")
while [[ $count != 2 ]]
do
sleep 1
@ -53,12 +49,10 @@ $CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 2, 'h
$CLICKHOUSE_CLIENT --query="SELECT count(*) from elog" # 5 rows
count=$($CLICKHOUSE_CLIENT --query="SELECT COUNT(*) FROM system.zookeeper where path = '/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/elog/s1/blocks'")
while [[ $count != 2 ]]
do
sleep 1
count=$($CLICKHOUSE_CLIENT --query="SELECT COUNT(*) FROM system.zookeeper where path = '/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/elog/s1/blocks'")
done
$CLICKHOUSE_CLIENT --query="INSERT INTO elog VALUES (toDate('2018-10-01'), 2, 'hello')"

20
tests/queries/0_stateless/01305_replica_create_drop_zookeeper.sh
View File

@ -8,21 +8,11 @@ set -e
function thread()
{
db_engine=`$CLICKHOUSE_CLIENT -q "SELECT engine FROM system.databases WHERE name='$CLICKHOUSE_DATABASE'"`
if [[ $db_engine == "Atomic" ]]; then
# Ignore "Replica already exists" exception
while true; do
$CLICKHOUSE_CLIENT -n -q "DROP TABLE IF EXISTS test_table_$1 NO DELAY;
CREATE TABLE test_table_$1 (a UInt8) ENGINE = ReplicatedMergeTree('/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/alter_table', 'r_$1') ORDER BY tuple();" 2>&1 |
grep -vP '(^$)|(^Received exception from server)|(^\d+\. )|because the last replica of the table was dropped right now|is already started to be removing by another replica right now|is already finished removing by another replica right now|Removing leftovers from table|Another replica was suddenly created|was successfully removed from ZooKeeper|was created by another server at the same moment|was suddenly removed|some other replicas were created at the same time|already exists'
done
else
while true; do
$CLICKHOUSE_CLIENT -n -q "DROP TABLE IF EXISTS test_table_$1;
CREATE TABLE test_table_$1 (a UInt8) ENGINE = ReplicatedMergeTree('/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/alter_table', 'r_$1') ORDER BY tuple();" 2>&1 |
grep -vP '(^$)|(^Received exception from server)|(^\d+\. )|because the last replica of the table was dropped right now|is already started to be removing by another replica right now|is already finished removing by another replica right now|Removing leftovers from table|Another replica was suddenly created|was successfully removed from ZooKeeper|was created by another server at the same moment|was suddenly removed|some other replicas were created at the same time'
done
fi
while true; do
$CLICKHOUSE_CLIENT -n -q "DROP TABLE IF EXISTS test_table_$1 SYNC;
CREATE TABLE test_table_$1 (a UInt8) ENGINE = ReplicatedMergeTree('/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/alter_table', 'r_$1') ORDER BY tuple();" 2>&1 |
grep -vP '(^$)|(^Received exception from server)|(^\d+\. )|because the last replica of the table was dropped right now|is already started to be removing by another replica right now|is already finished removing by another replica right now|Removing leftovers from table|Another replica was suddenly created|was successfully removed from ZooKeeper|was created by another server at the same moment|was suddenly removed|some other replicas were created at the same time'
done
}