NJrslv commented on code in PR #1629: URL: https://github.com/apache/cloudberry/pull/1629#discussion_r3015531338
########## gpcontrib/gp_stats_collector/src/UDSConnector.cpp: ########## @@ -0,0 +1,146 @@ +/*------------------------------------------------------------------------- + * + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + * + * UDSConnector.cpp + * + * IDENTIFICATION + * gpcontrib/gp_stats_collector/src/UDSConnector.cpp + * + *------------------------------------------------------------------------- + */ + +#include "UDSConnector.h" +#include "Config.h" +#include "GpscStat.h" +#include "log/LogOps.h" +#include "memory/gpdbwrappers.h" + +#include <chrono> +#include <string> +#include <sys/fcntl.h> +#include <sys/socket.h> +#include <sys/types.h> +#include <sys/un.h> +#include <thread> +#include <unistd.h> + +extern "C" { +#include "postgres.h" +} + +static void inline log_tracing_failure(const gpsc::SetQueryReq &req, + const std::string &event) +{ + ereport(LOG, (errmsg("Query {%d-%d-%d} %s tracing failed with error %m", + req.query_key().tmid(), req.query_key().ssid(), + req.query_key().ccnt(), event.c_str()))); +} + +bool +UDSConnector::report_query(const gpsc::SetQueryReq &req, + const std::string &event, const Config &config) +{ + sockaddr_un address{}; + address.sun_family = AF_UNIX; + const auto &uds_path = config.uds_path(); + + if (uds_path.size() >= sizeof(address.sun_path)) + { + ereport(WARNING, (errmsg("UDS path is too long for socket buffer"))); + GpscStat::report_error(); + return false; + } + strcpy(address.sun_path, uds_path.c_str()); + + const auto sockfd = socket(AF_UNIX, SOCK_STREAM, 0); + if (sockfd == -1) + { + log_tracing_failure(req, event); + GpscStat::report_error(); + return false; + } + + // Close socket automatically on error path. + struct SockGuard + { + int fd; + ~SockGuard() + { + close(fd); + } + } sock_guard{sockfd}; + + if (fcntl(sockfd, F_SETFL, O_NONBLOCK) == -1) + { + // That's a very important error that should never happen, so make it + // visible to an end-user and admins. + ereport(WARNING, + (errmsg("Unable to create non-blocking socket connection %m"))); + GpscStat::report_error(); + return false; + } + + if (connect(sockfd, reinterpret_cast<sockaddr *>(&address), + sizeof(address)) == -1) + { + log_tracing_failure(req, event); + GpscStat::report_bad_connection(); + return false; + } + + const auto data_size = req.ByteSizeLong(); + const auto total_size = data_size + sizeof(uint32_t); + auto *buf = static_cast<uint8_t *>(gpdb::palloc(total_size)); + // Free buf automatically on error path. + struct BufGuard + { + void *p; + ~BufGuard() + { + gpdb::pfree(p); + } + } buf_guard{buf}; + + *reinterpret_cast<uint32_t *>(buf) = data_size; + req.SerializeWithCachedSizesToArray(buf + sizeof(uint32_t)); + + int64_t sent = 0, sent_total = 0; + do + { + sent = send(sockfd, buf + sent_total, total_size - sent_total, + MSG_DONTWAIT); + if (sent > 0) + sent_total += sent; + } while (sent > 0 && size_t(sent_total) != total_size && + // the line below is a small throttling hack: + // if a message does not fit a single packet, we take a nap + // before sending the next one. + // Otherwise, MSG_DONTWAIT send might overflow the UDS + (std::this_thread::sleep_for(std::chrono::milliseconds(1)), true)); Review Comment: Partially done here - https://github.com/apache/cloudberry/pull/1629/changes/649da13fe7f06998d4e38e199884f67959ff2865. What I am thinking here is that we need to use `CHECK_FOR_INTERRUPTS()` along with `pg_usleep()` here, but `CHECK_FOR_INTERRUPTS()` internally might call `longjmp()` bypassing destructor call of C++ objects like `buf_guard` or `sock_guard` or maybe some other object on the stack, from some previous outer function. Using `WaitLatch()` seems like overkill for a 1ms sleep, so I just replaced `sleep_for()` with `pg_usleep()`. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
