Hongxu Ma created HAWQ-1627:
-------------------------------
Summary: Support setting the max protocol message size when
talking with HDFS
Key: HAWQ-1627
URL: https://issues.apache.org/jira/browse/HAWQ-1627
Project: Apache HAWQ
Issue Type: Improvement
Components: libhdfs
Reporter: Hongxu Ma
Assignee: Radar Lei
Fix For: 2.4.0.0-incubating
Now, the max size of protocol message in libhdfs is 64MB, and it cannot be
adjusted.
When the max size is reached (accessing a very big HDFS table/file), will see
the following lines in hawq master log:
{code}
2018-06-20 11:21:56.768003
CST,,,p75703,th-848100416,,,,0,,,seg-10000,,,,,"LOG","00000","3rd party error
log:
[libprotobuf ERROR google/protobuf/io/coded_stream.cc:208] A protocol message
was rejected because it was too big (more than 67108864 bytes). To increase
the limit (or to
disable these warnings), see CodedInputStream::SetTotalBytesLimit() in
google/protobuf/io/coded_stream.h.",,,,,,,,"SysLoggerMain","syslogger.c",518,
2018-06-20 11:21:56.771657
CST,,,p75703,th-848100416,,,,0,,,seg-10000,,,,,"LOG","00000","3rd party error
log:
2018-06-20 11:21:56.771492, p75751, th0x7fffcd7303c0, ERROR Failed to invoke
RPC call ""getFsStats"" on server ""localhost:9000"":
RpcChannel.cpp: 783: HdfsRpcException: RPC channel to ""localhost:9000"" got
protocol mismatch: RPC channel cannot parse response header.
@ Hdfs::Internal::RpcChannelImpl::readOneResponse(bool)
@ Hdfs::Internal::RpcChannelImpl::checkOneResponse()
@
Hdfs::Internal::RpcChannelImpl::invokeInternal(std::__1::shared_ptr<Hdfs::Internal::RpcRemoteCall>)
@
Hdfs::Internal::RpcChannelImpl:",,,,,,,,"SysLoggerMain","syslogger.c",518,
2018-06-20 11:21:56.771711
CST,,,p75703,th-848100416,,,,0,,,seg-10000,,,,,"LOG","00000","3rd party error
log:
:invoke(Hdfs::Internal::RpcCall const&)
@ Hdfs::Internal::NamenodeImpl::invoke(Hdfs::Internal::RpcCall const&)
@ Hdfs::Internal::NamenodeImpl::getFsStats()
@ Hdfs::Internal::NamenodeProxy::getFsStats()
@ Hdfs::Internal::FileSystemImpl::getFsStats()
@ Hdfs::Internal::FileSystemImpl::connect()
@ Hdfs::FileSystem::connect(char const*, char const*, char const*)
@ Hdfs::FileSystem::connect(char const*)
@ hdfsBuilderConnect
@ gpfs_hdfs_connect
@ HdfsConnect
@ HdfsGetConnection
@ HdfsGetFileBlockLocati",,,,,,,,"SysLoggerMain","syslogger.c",518,
{code}
Considering HDFS has a guc *"ipc.maximum.data.length"* to set it, HAWQ also
should has a guc about it.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
