thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-09-21 09:10:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

dbms: server: Retry after a timeout issue. [#METR-20825]

Browse Source
This commit is contained in:
Alexey Arno 2016-04-15 01:00:39 +03:00
parent b884690fbf
commit 68f427398d
1 changed files with 50 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
dbms/src/Storages/MergeTree/ReshardingWorker.cpp
View File

@ -697,13 +697,16 @@ void ReshardingWorker::perform(const std::string & job_descriptor, const std::st
}
else if (ex.code() == ErrorCodes::RESHARDING_REMOTE_NODE_UNAVAILABLE)
{
/// A remote performer has gone offline.
/// A remote performer has gone offline or we are experiencing network problems.
/// Put the current distributed job on hold. Also jab the job scheduler
/// so that it will come accross this distributed job even if no new jobs
/// are submitted.
setStatus(current_job.coordinator_id, getFQDNOrHostName(), STATUS_ON_HOLD,
ex.message());
dumped_coordinator_state = dumpCoordinatorState(current_job.coordinator_id);
if (current_job.isCoordinated())
{
setStatus(current_job.coordinator_id, getFQDNOrHostName(), STATUS_ON_HOLD,
ex.message());
dumped_coordinator_state = dumpCoordinatorState(current_job.coordinator_id);
}
softCleanup();
wakeUpTrackerThread();
}
@ -956,6 +959,20 @@ void ReshardingWorker::publishShardedPartitions()
pool.schedule([j, &tasks]{ tasks[j](); });
}
}
catch (const Poco::TimeoutException & ex)
{
try
{
pool.wait();
}
catch (...)
{
tryLogCurrentException(__PRETTY_FUNCTION__);
}
throw Exception{"Sharded partition upload operation timed out",
ErrorCodes::RESHARDING_REMOTE_NODE_UNAVAILABLE};
}
catch (...)
{
try
@ -1095,6 +1112,20 @@ void ReshardingWorker::commit()
}
}
}
catch (const Poco::TimeoutException & ex)
{
try
{
pool.wait();
}
catch (...)
{
tryLogCurrentException(__PRETTY_FUNCTION__);
}
throw Exception{"A remote operation timed out while committing",
ErrorCodes::RESHARDING_REMOTE_NODE_UNAVAILABLE};
}
catch (...)
{
try
@ -1260,6 +1291,21 @@ bool ReshardingWorker::checkAttachLogRecord(LogRecord & log_record)
pool.schedule([i, &tasks]{ tasks[i](); });
}
}
catch (const Poco::TimeoutException & ex)
{
try
{
pool.wait();
}
catch (...)
{
tryLogCurrentException(__PRETTY_FUNCTION__);
}
throw Exception{"Part checking on remote node timed out while attempting "
"to fix a failed ATTACH operation",
ErrorCodes::RESHARDING_REMOTE_NODE_UNAVAILABLE};
}
catch (...)
{
try