h2load: add an option to write per-request logs
This commit is contained in:
parent
eb6792533d
commit
d1b3a83f59
|
@ -748,6 +748,7 @@ void Client::on_header(int32_t stream_id, const uint8_t *name, size_t namelen,
|
|||
}
|
||||
}
|
||||
|
||||
stream.req_stat.status = status;
|
||||
if (status >= 200 && status < 300) {
|
||||
++worker->stats.status[2];
|
||||
stream.status_success = 1;
|
||||
|
@ -775,6 +776,7 @@ void Client::on_status_code(int32_t stream_id, uint16_t status) {
|
|||
return;
|
||||
}
|
||||
|
||||
stream.req_stat.status = status;
|
||||
if (status >= 200 && status < 300) {
|
||||
++worker->stats.status[2];
|
||||
stream.status_success = 1;
|
||||
|
@ -821,6 +823,21 @@ void Client::on_stream_close(int32_t stream_id, bool success, bool final) {
|
|||
}
|
||||
++worker->stats.req_done;
|
||||
++req_done;
|
||||
|
||||
if (worker->log) {
|
||||
auto start = std::chrono::duration_cast<std::chrono::microseconds>(
|
||||
req_stat->request_wall_time.time_since_epoch());
|
||||
auto delta = std::chrono::duration_cast<std::chrono::microseconds>(
|
||||
req_stat->stream_close_time - req_stat->request_time);
|
||||
*worker->log << start.count() << '\t' << (success ? req_stat->status : -1)
|
||||
<< '\t' << delta.count() << '\n';
|
||||
// Flushing manually is important to ensure atomicity of lines, but
|
||||
// doing it after each line (e.g. with std::endl) is noticeably slow.
|
||||
if (++worker->log_pending == 16) {
|
||||
worker->log->flush();
|
||||
worker->log_pending = 0;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
worker->report_progress();
|
||||
|
@ -1177,6 +1194,7 @@ int Client::write_tls() {
|
|||
|
||||
void Client::record_request_time(RequestStat *req_stat) {
|
||||
req_stat->request_time = std::chrono::steady_clock::now();
|
||||
req_stat->request_wall_time = std::chrono::system_clock::now();
|
||||
}
|
||||
|
||||
void Client::record_connect_start_time() {
|
||||
|
@ -1246,7 +1264,8 @@ Worker::Worker(uint32_t id, SSL_CTX *ssl_ctx, size_t req_todo, size_t nclients,
|
|||
nreqs_rem(req_todo % nclients),
|
||||
rate(rate),
|
||||
max_samples(max_samples),
|
||||
next_client_id(0) {
|
||||
next_client_id(0),
|
||||
log_pending(0) {
|
||||
if (!config->is_rate_mode() && !config->is_timing_based_mode()) {
|
||||
progress_interval = std::max(static_cast<size_t>(1), req_todo / 10);
|
||||
} else {
|
||||
|
@ -1281,6 +1300,11 @@ Worker::Worker(uint32_t id, SSL_CTX *ssl_ctx, size_t req_todo, size_t nclients,
|
|||
} else {
|
||||
current_phase = Phase::MAIN_DURATION;
|
||||
}
|
||||
|
||||
if (!config->log_file.empty()) {
|
||||
log = std::make_unique<std::ofstream>(
|
||||
config->log_file, std::ios_base::out | std::ios_base::app);
|
||||
}
|
||||
}
|
||||
|
||||
Worker::~Worker() {
|
||||
|
@ -1940,6 +1964,14 @@ Options:
|
|||
this option value and the value which server specified.
|
||||
Default: )"
|
||||
<< util::utos_unit(config.encoder_header_table_size) << R"(
|
||||
--log-file=<PATH>
|
||||
Write per-request information to a file as tab-separated
|
||||
columns: start time as microseconds since epoch; HTTP
|
||||
status code; microseconds until end of response. More
|
||||
columns may be added later. Rows are ordered by end-of-
|
||||
response time when using one worker thread, but may
|
||||
appear slightly out of order with multiple threads due
|
||||
to buffering. Status code is -1 for failed streams.
|
||||
-v, --verbose
|
||||
Output debug information.
|
||||
--version Display version information and exit.
|
||||
|
@ -1996,6 +2028,7 @@ int main(int argc, char **argv) {
|
|||
{"header-table-size", required_argument, &flag, 7},
|
||||
{"encoder-header-table-size", required_argument, &flag, 8},
|
||||
{"warm-up-time", required_argument, &flag, 9},
|
||||
{"log-file", required_argument, &flag, 10},
|
||||
{nullptr, 0, nullptr, 0}};
|
||||
int option_index = 0;
|
||||
auto c = getopt_long(argc, argv,
|
||||
|
@ -2219,6 +2252,15 @@ int main(int argc, char **argv) {
|
|||
exit(EXIT_FAILURE);
|
||||
}
|
||||
break;
|
||||
case 10:
|
||||
// --log-file
|
||||
std::ofstream out(optarg, std::ios_base::out | std::ios_base::app);
|
||||
if (!out.is_open()) {
|
||||
std::cerr << "--log-file: cannot write to " << optarg << std::endl;
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
config.log_file = optarg;
|
||||
break;
|
||||
}
|
||||
break;
|
||||
default:
|
||||
|
|
|
@ -71,6 +71,7 @@ struct Config {
|
|||
std::string host;
|
||||
std::string ifile;
|
||||
std::string ciphers;
|
||||
std::string log_file;
|
||||
// length of upload data
|
||||
int64_t data_length;
|
||||
addrinfo *addrs;
|
||||
|
@ -122,10 +123,14 @@ struct Config {
|
|||
struct RequestStat {
|
||||
// time point when request was sent
|
||||
std::chrono::steady_clock::time_point request_time;
|
||||
// same, but in wall clock reference frame
|
||||
std::chrono::system_clock::time_point request_wall_time;
|
||||
// time point when stream was closed
|
||||
std::chrono::steady_clock::time_point stream_close_time;
|
||||
// upload data length sent so far
|
||||
int64_t data_offset;
|
||||
// HTTP status code
|
||||
int status;
|
||||
// true if stream was successfully closed. This means stream was
|
||||
// not reset, but it does not mean HTTP level error (e.g., 404).
|
||||
bool completed;
|
||||
|
@ -266,6 +271,10 @@ struct Worker {
|
|||
// specified
|
||||
ev_timer duration_watcher;
|
||||
ev_timer warmup_watcher;
|
||||
// Stream to write per-request stats to.
|
||||
std::unique_ptr<std::ostream> log;
|
||||
// Number of unflushed lines in the log.
|
||||
uint32_t log_pending;
|
||||
|
||||
Worker(uint32_t id, SSL_CTX *ssl_ctx, size_t nreq_todo, size_t nclients,
|
||||
size_t rate, size_t max_samples, Config *config);
|
||||
|
|
Loading…
Reference in New Issue