mirrors/dragonfly
1
0
Fork 0
mirror of https://github.com/dragonflydb/dragonfly.git synced 2025-05-11 18:35:46 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

feat(server_family): Add backup/restore Prometheus metrics (#3520)

Browse source 
* feat(server_family): Add backup/restore Prometheus metrics

fixes dragonflydb#3210

---------

Signed-off-by: Stepan Bagritsevich <bagr.stepan@gmail.com>
This commit is contained in:
Stepan Bagritsevich 2024-08-23 23:36:31 +02:00 • committed by GitHub
parent a66cbe150e
commit 80c3579596
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 97 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

35
src/server/server_family.cc
View file

@ -869,8 +869,14 @@ void ServerFamily::Init(util::AcceptServer* acceptor, std::vector<facade::Listen
}
void ServerFamily::LoadFromSnapshot() {
{
std::lock_guard lk{loading_stats_mu_};
loading_stats_.restore_count++;
}
const auto load_path_result =
snapshot_storage_->LoadPath(GetFlag(FLAGS_dir), GetFlag(FLAGS_dbfilename));
if (load_path_result) {
const std::string load_path = *load_path_result;
if (!load_path.empty()) {
@ -880,6 +886,8 @@ void ServerFamily::LoadFromSnapshot() {
if (std::error_code(load_path_result.error()) == std::errc::no_such_file_or_directory) {
LOG(WARNING) << "Load snapshot: No snapshot found";
} else {
std::lock_guard lk{loading_stats_mu_};
loading_stats_.failed_restore_count++;
LOG(ERROR) << "Failed to load snapshot: " << load_path_result.error().Format();
}
}
@ -904,7 +912,13 @@ void ServerFamily::Shutdown() {
if (save_on_shutdown_ && !absl::GetFlag(FLAGS_dbfilename).empty()) {
shard_set->pool()->GetNextProactor()->Await([this] {
if (GenericError ec = DoSave(); ec) {
GenericError ec = DoSave();
std::lock_guard lk{loading_stats_mu_};
loading_stats_.backup_count++;
if (ec) {
loading_stats_.failed_backup_count++;
LOG(WARNING) << "Failed to perform snapshot " << ec.Format();
}
});
@ -1061,7 +1075,12 @@ void ServerFamily::SnapshotScheduling() {
};
GenericError ec = DoSave();
std::lock_guard lk{loading_stats_mu_};
loading_stats_.backup_count++;
if (ec) {
loading_stats_.failed_backup_count++;
LOG(WARNING) << "Failed to perform snapshot " << ec.Format();
}
}
@ -1270,6 +1289,15 @@ void PrintPrometheusMetrics(const Metrics& m, DflyCmd* dfly_cmd, StringResponse*
AppendMetricWithoutLabels("lua_blocked_total", "", m.lua_stats.blocked_cnt, MetricType::COUNTER,
&resp->body());
AppendMetricWithoutLabels("backups_total", "", m.loading_stats.backup_count, MetricType::COUNTER,
&resp->body());
AppendMetricWithoutLabels("failed_backups_total", "", m.loading_stats.failed_backup_count,
MetricType::COUNTER, &resp->body());
AppendMetricWithoutLabels("restores_total", "", m.loading_stats.restore_count,
MetricType::COUNTER, &resp->body());
AppendMetricWithoutLabels("failed_restores_total", "", m.loading_stats.failed_restore_count,
MetricType::COUNTER, &resp->body());
// Net metrics
AppendMetricWithoutLabels("net_input_bytes_total", "", conn_stats.io_read_bytes,
MetricType::COUNTER, &resp->body());
@ -2035,6 +2063,11 @@ Metrics ServerFamily::GetMetrics(Namespace* ns) const {
}
}
{
std::lock_guard lk{loading_stats_mu_};
result.loading_stats = loading_stats_;
}
// Update peak stats. We rely on the fact that GetMetrics is called frequently enough to
// update peak_stats_ from it.
lock_guard lk{peak_stats_mu_};