Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix: node qps sum inconsistent with app qps sum #511

Merged
merged 4 commits into from
Apr 9, 2020
Merged

fix: node qps sum inconsistent with app qps sum #511

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5a69322
fix bug and add cu
hycdong Apr 7, 2020
b4c8fc9
small fix
hycdong Apr 7, 2020
57f0348
Merge branch 'master' into fix_qps
acelyc111 Apr 8, 2020
3896516
Merge branch 'master' into fix_qps
hycdong Apr 8, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion src/shell/commands.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,8 @@ struct list_nodes_helper
double put_p99;
double multi_get_p99;
double multi_put_p99;
double read_cu;
double write_cu;
list_nodes_helper(const std::string &n, const std::string &s)
: node_name(n),
node_status(s),
Expand All @@ -73,7 +75,9 @@ struct list_nodes_helper
get_p99(0.0),
put_p99(0.0),
multi_get_p99(0.0),
multi_put_p99(0.0)
multi_put_p99(0.0),
read_cu(0.0),
write_cu(0.0)
{
}
};
Expand Down
98 changes: 77 additions & 21 deletions src/shell/commands/node_management.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,7 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
bool resolve_ip = false;
bool resource_usage = false;
bool show_qps = false;
bool show_latency = false;
bool json = false;
optind = 0;
while (true) {
Expand All @@ -81,6 +82,7 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
break;
case 'q':
show_qps = true;
show_latency = true;
break;
case 'j':
json = true;
Expand Down Expand Up @@ -240,13 +242,67 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
return true;
}

// TODO(heyuchen): add cu statistics
::dsn::command command;
command.cmd = "perf-counters-by-prefix";
command.arguments.push_back("replica*app.pegasus*get_qps");
command.arguments.push_back("replica*app.pegasus*multi_get_qps");
command.arguments.push_back("replica*app.pegasus*put_qps");
command.arguments.push_back("replica*app.pegasus*multi_put_qps");
command.arguments.push_back("replica*app.pegasus*recent.read.cu");
command.arguments.push_back("replica*app.pegasus*recent.write.cu");

std::vector<std::pair<bool, std::string>> results;
call_remote_command(sc, nodes, command, results);

for (int i = 0; i < nodes.size(); ++i) {
dsn::rpc_address node_addr = nodes[i].address;
auto tmp_it = tmp_map.find(node_addr);
if (tmp_it == tmp_map.end())
continue;
if (!results[i].first) {
std::cout << "query perf counter info from node " << node_addr.to_string()
<< " failed" << std::endl;
return true;
}
dsn::perf_counter_info info;
dsn::blob bb(results[i].second.data(), 0, results[i].second.size());
if (!dsn::json::json_forwarder<dsn::perf_counter_info>::decode(bb, info)) {
std::cout << "decode perf counter info from node " << node_addr.to_string()
<< " failed, result = " << results[i].second << std::endl;
return true;
}
if (info.result != "OK") {
std::cout << "query perf counter info from node " << node_addr.to_string()
<< " returns error, error = " << info.result << std::endl;
return true;
}
list_nodes_helper &h = tmp_it->second;
for (dsn::perf_counter_metric &m : info.counters) {
if (m.name.find("replica*app.pegasus*get_qps") != std::string::npos)
h.get_qps += m.value;
else if (m.name.find("replica*app.pegasus*multi_get_qps") != std::string::npos)
h.multi_get_qps += m.value;
else if (m.name.find("replica*app.pegasus*put_qps") != std::string::npos)
h.put_qps += m.value;
else if (m.name.find("replica*app.pegasus*multi_put_qps") != std::string::npos)
h.multi_put_qps += m.value;
else if (m.name.find("replica*app.pegasus*recent.read.cu") != std::string::npos)
h.read_cu += m.value;
else if (m.name.find("replica*app.pegasus*recent.write.cu") != std::string::npos)
h.write_cu += m.value;
}
}
}

if (show_latency) {
std::vector<node_desc> nodes;
if (!fill_nodes(sc, "replica-server", nodes)) {
std::cout << "get replica server node list failed" << std::endl;
return true;
}

::dsn::command command;
command.cmd = "perf-counters-by-postfix";
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_GET.qps");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_PUT.qps");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_MULTI_GET.qps");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_MULTI_PUT.qps");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_GET.latency.server");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_PUT.latency.server");
command.arguments.push_back("zion*profiler*RPC_RRDB_RRDB_MULTI_GET.latency.server");
Expand Down Expand Up @@ -278,15 +334,7 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
}
list_nodes_helper &h = tmp_it->second;
for (dsn::perf_counter_metric &m : info.counters) {
if (m.name.find("RPC_RRDB_RRDB_GET.qps") != std::string::npos)
h.get_qps = m.value;
else if (m.name.find("RPC_RRDB_RRDB_PUT.qps") != std::string::npos)
h.put_qps = m.value;
else if (m.name.find("RPC_RRDB_RRDB_MULTI_GET.qps") != std::string::npos)
h.multi_get_qps = m.value;
else if (m.name.find("RPC_RRDB_RRDB_MULTI_PUT.qps") != std::string::npos)
h.put_qps = m.value;
else if (m.name.find("RPC_RRDB_RRDB_GET.latency.server") != std::string::npos)
if (m.name.find("RPC_RRDB_RRDB_GET.latency.server") != std::string::npos)
h.get_p99 = m.value;
else if (m.name.find("RPC_RRDB_RRDB_PUT.latency.server") != std::string::npos)
h.put_p99 = m.value;
Expand Down Expand Up @@ -328,12 +376,16 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
}
if (show_qps) {
tp.add_column("get_qps", tp_alignment::kRight);
tp.add_column("get_p99(ms)", tp_alignment::kRight);
tp.add_column("mget_qps", tp_alignment::kRight);
tp.add_column("mget_p99(ms)", tp_alignment::kRight);
tp.add_column("read_cu", tp_alignment::kRight);
tp.add_column("put_qps", tp_alignment::kRight);
tp.add_column("put_p99(ms)", tp_alignment::kRight);
tp.add_column("mput_qps", tp_alignment::kRight);
tp.add_column("write_cu", tp_alignment::kRight);
}
if (show_latency) {
tp.add_column("get_p99(ms)", tp_alignment::kRight);
tp.add_column("mget_p99(ms)", tp_alignment::kRight);
tp.add_column("put_p99(ms)", tp_alignment::kRight);
tp.add_column("mput_p99(ms)", tp_alignment::kRight);
}
for (auto &kv : tmp_map) {
Expand All @@ -354,13 +406,17 @@ bool ls_nodes(command_executor *e, shell_context *sc, arguments args)
}
if (show_qps) {
tp.append_data(kv.second.get_qps);
tp.append_data(kv.second.get_p99 / 1000000);
tp.append_data(kv.second.multi_get_qps);
tp.append_data(kv.second.multi_get_p99 / 1000000);
tp.append_data(kv.second.read_cu);
tp.append_data(kv.second.put_qps);
tp.append_data(kv.second.put_p99 / 1000000);
tp.append_data(kv.second.multi_put_qps);
tp.append_data(kv.second.multi_put_p99 / 1000000);
tp.append_data(kv.second.write_cu);
}
if (show_latency) {
tp.append_data(kv.second.get_p99 / 1e6);
tp.append_data(kv.second.multi_get_p99 / 1e6);
tp.append_data(kv.second.put_p99 / 1e6);
tp.append_data(kv.second.multi_put_p99 / 1e6);
}
}
mtp.add(std::move(tp));
Expand Down