Compare commits

...

17 commits

Author SHA1 Message Date
5f4765d196 Update of README 2023-11-22 11:18:36 +01:00
c0f6adcf67 Remove debug output and fix whitespace 2023-11-22 11:03:56 +01:00
035803fbe2 Error handling for jobid without data 2023-11-22 10:52:49 +01:00
ee106619ea Outline main into class 2023-11-22 10:30:36 +01:00
36bfb06228 Add factory function for power from data base 2023-11-22 09:01:05 +01:00
f9ebd68654 Refactor database class for multiple querries 2023-11-20 16:50:22 +01:00
9294b04291 Introduce class for database handling 2023-11-20 16:23:36 +01:00
bacc9e63c9 Use config object 2023-11-20 16:19:34 +01:00
8596fb4184 Remove/comment debug output 2023-11-20 16:04:19 +01:00
f90b3fc1da Rearrange class methods 2023-11-20 15:53:52 +01:00
3a8170e7c9 Add missing import statement 2023-11-20 11:39:11 +01:00
294a4e1273 Add argument parser 2023-11-17 16:25:51 +01:00
b54e31f4fd Lazy import of sqlalchemy 2023-11-17 16:24:00 +01:00
3d52b1b2c7 Format README 2023-11-17 16:20:57 +01:00
831a85639b Add prototype to query job power from database 2023-10-26 16:19:16 +02:00
8a60c2dd9f Merge pull request 'Utility to fetch from psql database and export to csv like format' (#1) from hpckhald/hawk-utils-scripts:pgsql-db into pgsql-db
Reviewed-on: #1
2023-09-14 12:18:49 +00:00
b321a25fb0 First commit for getting data from PG-SQL 2023-08-09 17:39:31 +02:00
4 changed files with 434 additions and 0 deletions

View file

@ -1,3 +1,22 @@
# hawk-utils-scripts/monitoring/db
Use this directory for scripts, etc, that operate on or extract data from the the database.
## scripts/get_detailed_power.py
Python script which querries the database for a list of jobids and produces a file for each jobid.
Those files contain time-resolved power consumption data and are meant to be consumed by the utility in [monitoring/logs/scripts/plot_energy_logs.sh](../logs/README.md#scripts/plot_energy_logs.sh).
Requirements:
- Python module sqlalchemy `python -m pip install sqlalchemy`
Usage:
```bash
ssh monitoring
./get_detailed_power.py 2335001 7678069.hawk-pbs5
Created file detailed_power.2335001.hawk-pbs5.100000-1000004.csv
Created file detailed_power.7678069.hawk-pbs5.2432345-423423.csv
```

View file

@ -0,0 +1,8 @@
#!/bin/bash
module load gcc/13.1.0
module list
PG_DIR=/quobyte/qb1/hpc34576/hpc34576/postgres
echo "Before running, use \"export LD_LIBRARY_PATH=/quobyte/qb1/hpc34576/hpc34576/postgres/lib:\$LD_LIBRARY_PATH\""
gcc -O3 -I${PG_DIR}/include -L${PG_DIR}/lib poc.c -lpq

View file

@ -0,0 +1,195 @@
/* License? */
#include"libpq-fe.h"
#include<stdio.h>
#include<stdlib.h>
#include<string.h>
#define MIN(a,b) ((a)<(b)?(a):(b))
static int GetNumNodes(PGconn *const conn, const char *const jobid)
{
int n= 0;
char cmd[84];
sprintf(cmd, "select node_amount from jobs where job_id='%7s.hawk-pbs5'", jobid);
PGresult *res= PQexec(conn, cmd);
if(PGRES_TUPLES_OK!= PQresultStatus(res)) {
goto bye;
}
const int nfields= PQnfields(res);
const int ntups= PQntuples(res);
if(1!= nfields|| 1!= ntups) {
goto bye;
}
n= atoi(PQgetvalue(res, 0, 0));
bye:
PQclear(res);
return n;
}
static int NumNodes= 0;
static char (*NodeNames)[32]= NULL;
inline static int GetNodeIx(const char *const name)
{
int i;
for(i= 0; i< NumNodes; ++i) {
if(strlen(NodeNames[i])> 0) {
if(0== strcmp(NodeNames[i], name)) {
break;
}
} else {
strcpy(NodeNames[i], name);
break;
}
}
if(i>= NumNodes) {
printf("Strange situation with node-name!\n");
}
return i;
}
inline static void WriteTimeVal(FILE *const fp, const long long tp, const double *const vals, const int n)
{
fprintf(fp, " %lld", tp);
for(int i= 0; i< n; ++i) {
fprintf(fp, " %.1lf", vals[i]);
}
fprintf(fp, "\n");
}
/* Somebody took time to write this query at <https://kb.hlrs.de/monitoring/index.php/TimescaleDB#Job_data> */
/* Used the query as is instead of building the query as the database code is expected to be most optimised */
char QString[1170]=
"with job as (\n"
" select job_id, starttime, endtime, nodes from jobs where job_id='xxxxxxx.hawk-pbs5'\n"
"),\n"
"node_series as (\n"
" select n.name, scmcavg.id as series_id from nodes n\n"
" inner join (select * from label where key='node') l on n.id = l.value::int\n"
" inner join series_cmc_power_racktraynodepoweravg scmcavg on l.id = scmcavg.labels[(\n"
" select pos from label_key_position\n"
" where metric_category= 'cmc_power'\n"
" and metric_name = 'RackTrayNodePowerAvg'\n"
" and key = 'node'\n"
" )]\n"
" where n.id = any((select nodes from job)::int[])\n"
")\n"
"select a.time, a.value, ns.name from (\n"
" select\n"
" time_bucket(extract ('epoch' from '5 seconds'::interval)::int*1000, cmcavg.ts) as time,\n"
" cmcavg.series_id::varchar,\n"
" avg(cmcavg.val) AS value\n"
" from cmc_power_racktraynodepoweravg cmcavg\n"
"where\n"
" ts <= (select endtime from job)\n"
" and ts >= (select starttime from job)\n"
" and series_id = Any(select series_id from node_series)\n"
"group by time, cmcavg.series_id order by time desc) a\n"
"inner join node_series ns on a.series_id::int = ns.series_id;\n";
int main(int argc, char *argv[])
{
if(argc< 2|| 7!= strlen(argv[1])|| 0== atoi(argv[1])) {
printf("Invalid job-id (%s)\n", argc< 2? "(null)": argv[1]);
return 0;
}
PGconn *conn= PQsetdbLogin("hawk-monitor4", "5432", "", "", "coe_mon", "hpc", "");
if(CONNECTION_OK!= PQstatus(conn)) {
printf("Cannot connect to the database!\n");
return 0;
}
NumNodes= GetNumNodes(conn, argv[1]);
if(0== NumNodes) {
printf("Cannot retrieve number of nodes for the job-id: %d!\n", argv[1]);
goto bye;
}
char *ptr= strstr(QString, "xxxxxxx.hawk-pbs5");
strncpy(ptr, argv[1], 7);
PGresult *res= PQexec(conn, QString);
if(PGRES_TUPLES_OK!= PQresultStatus(res)) {
printf("Problem here!\n");
PQclear(res);
}
NodeNames= (char (*)[32]) malloc(sizeof(char[32])* NumNodes);
for(int i= 0; i< NumNodes; ++i) {
memset(NodeNames[i], 0, sizeof(char[32]));
}
const int nfields= PQnfields(res);
/* printf("%d fields\n", nfields); */
int itime= -1, inname= -1, ival= -1;
for(int i= 0; i< nfields; ++i) {
const char *fieldname= PQfname(res, i);
if(0== strcmp("time", fieldname)) {
itime= i;
} else if(0== strcmp("name", fieldname)) {
inname= i;
} else if(0== strcmp("value", fieldname)) {
ival= i;
} else {
printf("Do not know how to read field \"%s\"\n", fieldname);
}
}
if(-1== itime|| -1== inname|| -1== ival) {
printf("Unknown fields causing program to exit!\n");
PQclear(res);
goto bye;
}
/* printf("itime= %d, inname= %d, ival= %d\n", itime, inname, ival); */
int ntups= PQntuples(res);
/* printf("%d tuples\n", ntups); */
if(0!= ntups% NumNodes) {
printf("#entries (%d) is not integral multiple of #nodes (%d)\n", ntups, NumNodes);
}
int tup0= 0, tup1= ntups, tupstep= 1;
long long tp_min= atoll(PQgetvalue(res, tup0, itime));
if(tp_min> atoll(PQgetvalue(res, ntups- 1, itime))) {
printf("Time in reverse order!\n");
tup0= ntups- 1; tup1= -1; tupstep= -1;
tp_min= atoll(PQgetvalue(res, tup0, itime));
}
long long tp, last_tp= tp_min;
double *vals= malloc(sizeof(double)* NumNodes);
memset(vals, 0, sizeof(double)* NumNodes);
char nname[32];
FILE *fp= fopen("oldstyleoutput.dat", "w");
for(int it= tup0; tup1!= it; it+= tupstep) {
tp= atoll(PQgetvalue(res, it, itime)); /* - tp_min; */
if(tp!= last_tp&& tp< last_tp) { /* some sanity check because I don't understand how the database querry works! */
printf("Strange ordering!\n");
}
if(tp> last_tp) { /* write out the last time-point */
WriteTimeVal(fp, last_tp, vals, NumNodes);
memset(vals, 0, sizeof(double)* NumNodes);
last_tp= tp;
}
int nodeix= GetNodeIx(PQgetvalue(res, it, inname));
vals[nodeix]= atof(PQgetvalue(res, it, ival));
}
WriteTimeVal(fp, last_tp, vals, NumNodes);
fprintf(fp, "#offsetted-time-point (ms)");
for(int i= 0; i< NumNodes; ++i) {
fprintf(fp, " %s", NodeNames[i]);
}
fprintf(fp, "\n");
fclose(fp); fp= NULL;
if(NULL!= vals) { free(vals); vals= NULL; }
if(NULL!= NodeNames) { free(NodeNames); NodeNames= NULL; }
PQclear(res);
bye:
PQfinish(conn);
return 0;
}

View file

@ -0,0 +1,212 @@
import argparse
import numpy as np
from collections import OrderedDict
import os.path
def parse_arguments():
parser = argparse.ArgumentParser(
description='Produce detailed power usage data for a list of jobids.')
parser.add_argument('-v', '--verbose', action='store_true',
help='Show database querries, etc.')
parser.add_argument('jobid', type=parse_jobid,
nargs='+',
help='Job ID such as "2260215" or "2260215.hawk-pbs5"')
return parser.parse_args()
def parse_jobid(s):
import re
hawkpbs = r'.hawk-pbs5'
jobid = re.sub(hawkpbs, '', s)
if not jobid.isdigit():
raise argparse.ArgumentTypeError(f'invalid job ID "{s}"')
return jobid
class Power:
def __init__(self, nodes):
self.nodes = nodes
self.epochs = OrderedDict()
self.first_ts = None
self.last_ts = None
@classmethod
def from_list(cls, data):
"""Assumes data is a list of tuples (timestamp, value, node)"""
nodes = list(set([line[2] for line in data]))
cls = Power(nodes)
# for now ignore order to nodes
values = {}
for l in data:
ts = l[0]
if ts not in values:
values[ts] = []
# node = l[1]
power = l[1]
values[ts].append(power)
epochs = sorted(values.keys())
for epoch in epochs:
cls.insert_epoch(epoch, values[epoch])
return cls
@classmethod
def from_db(cls, db, jobid, interval):
all_list = db.db_to_list(jobid, interval)
if not all_list:
raise RuntimeError
return Power.from_list(all_list)
def to_file(self, jobid):
"""Dumps power data to file. Returns filename is succesfull and None if unsucessfull."""
fname = self.filename(jobid)
if os.path.exists(fname):
print("Error: cowardly refusing to overwrite file ", fname)
return None
try:
with open(fname, "w+") as f:
f.write(self.header())
f.write(self.body())
except IOError:
print("Error: could not write to file ", filename)
fname = None
return fname
def insert_epoch(self, ts, values):
self.epochs[ts] = values
if not self.first_ts:
self.first_ts = ts
self.last_ts = ts
def header(self):
hd = "# all timestamp have unit miliseconds since unix epoch\n"
hd += "# all power values have unit Watt\n"
hd += "timestamp,delta_t,head_node_power,avg_node_power,median_node_power,min_node_power,max_node_power,std_dev_node_power"
# add node names here instead
hd += ",NO_NODE_NAMES_YET\n"
return hd
def body(self):
_body = ""
for epoch in self.epochs.items():
_body += Power.pretty_print(self.summarize_epoch(epoch))
return _body
@staticmethod
def summarize_time(ts):
return ts, -1
@staticmethod
def summarize_values(val):
values = np.asarray(val)
head = values[0]
min, max = values.min(), values.max()
avg, stddev = values.mean(), values.std()
median = np.median(values)
return head, avg, median, min, max, stddev
@staticmethod
def summarize_epoch(epoch):
ts, values = epoch
return Power.summarize_time(ts) \
+ Power.summarize_values(values)
# + values
@staticmethod
def pretty_print(args):
return ",".join(str(a) for a in args) + '\n'
def filename(self, jobid):
fname = "detailed_power_{jobid}.hawk-pbs5.{first}-{last}.csv".format(
jobid=jobid, first=self.first_ts, last=self.last_ts
)
return fname
class MonitoringDB:
def __init__(self, verbose):
self.connection = self.init_db(verbose)
@staticmethod
def init_db(verbose):
import sqlalchemy as db
_verbose = False #True
engine = db.create_engine('postgresql://hpc@hawk-monitor4:5432/coe_mon', echo=verbose)
connection = engine.connect()
return connection
def close_db(self):
return self.connection.close()
@staticmethod
def build_query(jobid, interval):
import sqlalchemy as db
query_string = """with job as (
select job_id, starttime, endtime, nodes from jobs where job_id='{jobid}.hawk-pbs5'
),
node_series as(
select n.name, scmcavg.id as series_id from nodes n
inner join (select * from label where key='node') l on n.id = l.value::int
inner join series_cmc_power_racktraynodepoweravg scmcavg on l.id = scmcavg.labels[(
select pos from label_key_position
where metric_category= 'cmc_power'
and metric_name = 'RackTrayNodePowerAvg'
and key = 'node'
)]
where n.id = any((select nodes from job)::int[])
)
select a.time, a.value, ns.name from (
select
time_bucket(extract ('epoch' from '{interval} seconds'::interval)::int*1000, cmcavg.ts) as time,
cmcavg.series_id::varchar,
avg(cmcavg.val) AS value
from cmc_power_racktraynodepoweravg cmcavg
where
ts <= (select endtime from job)
and ts >= (select starttime from job)
and series_id = Any(select series_id from node_series)
group by time, cmcavg.series_id order by time desc) a
inner join node_series ns on a.series_id::int = ns.series_id;
"""
return db.text(query_string.format(jobid=jobid, interval=interval))
def db_to_list(self, jobid, interval):
query = self.build_query(jobid, interval)
return self.connection.execute(query).fetchall()
def db_to_pf(self, jobid, inerval):
query = self.build_query(jobid, interval)
return pd.read_sql(query, con=self.connection)
class App:
def __init__(self, config):
self.config = config
self.db = MonitoringDB(self.config.verbose)
def run_all(self):
for jobid in self.config.jobid:
try:
power = Power.from_db(self.db, jobid, self.config.interval)
except RuntimeError:
print('No data found for job ID "{}"'.format(jobid))
continue
fn = power.to_file(jobid)
if fn:
print('Created file {fn}'.format(fn=fn))
if __name__ == "__main__":
config = parse_arguments()
config.interval = 5
main = App(config)
main.run_all()