Come misuro i tempi di richiesta e risposta contemporaneamente usando cURL?

Ho un servizio Web che riceve i dati in formato JSON, li elabora e quindi restituisce il risultato al richiedente.

Voglio misurare la richiesta, la risposta e il tempo totale usando cURL.

La mia richiesta di esempio è simile a:

curl -X POST -d @file server:port

e attualmente lo misuro usando il timecomando in Linux:

time curl -X POST -d @file server:port

Tuttavia, il comando time misura solo il tempo totale , il che non è proprio quello che sto cercando.

Esiste un modo per misurare i tempi di richiesta e risposta utilizzando cURL?

Da questo brillante post sul blog ... https://blog.josephscott.org/2011/10/14/timing-details-with-curl/

cURL supporta l'output formattato per i dettagli della richiesta (vedere la manpage cURL per i dettagli , sotto -w, –write-out <format>). Per i nostri scopi ci concentreremo solo sui dettagli di temporizzazione forniti. I tempi sotto sono in secondi .

1. Crea un nuovo file, curl-format.txt e incollalo:

       time_namelookup:  %{time_namelookup}s\n
          time_connect:  %{time_connect}s\n
       time_appconnect:  %{time_appconnect}s\n
      time_pretransfer:  %{time_pretransfer}s\n
         time_redirect:  %{time_redirect}s\n
    time_starttransfer:  %{time_starttransfer}s\n
                       ----------\n
            time_total:  %{time_total}s\n
   

2. Fare una richiesta:

   curl -w "@curl-format.txt" -o /dev/null -s "http://wordpress.com/"
   

   O su Windows, è ...

   curl -w "@curl-format.txt" -o NUL -s "http://wordpress.com/"
   

Cosa fa questo:

-w "@curl-format.txt"dice a cURL di usare il nostro file di formato
-o /dev/nullreindirizza l'output della richiesta a / dev / null
-s dice a cURL di non mostrare un indicatore di progresso
"http://wordpress.com/"è l'URL che stiamo richiedendo. Utilizza le virgolette in particolare se l'URL ha i parametri "&" della stringa di query

Ed ecco cosa ottieni:

   time_namelookup:  0.001s
      time_connect:  0.037s
   time_appconnect:  0.000s
  time_pretransfer:  0.037s
     time_redirect:  0.000s
time_starttransfer:  0.092s
                   ----------
        time_total:  0.164s

Crea un collegamento Linux / Mac (alias)

alias curltime="curl -w \"@$HOME/.curl-format.txt\" -o NUL -s "

Quindi puoi semplicemente chiamare ...

curltime wordpress.org

Grazie al commentatore Pete Doyle!

Crea uno script autonomo Linux / Mac

Questo script non richiede un file .txt separato per contenere la formattazione.

Crea un nuovo file, curltime, da qualche parte nel tuo percorso eseguibile e incollalo:

#!/bin/bash

curl -w @- -o /dev/null -s "$@" <<'EOF'
    time_namelookup:  %{time_namelookup}\n
       time_connect:  %{time_connect}\n
    time_appconnect:  %{time_appconnect}\n
   time_pretransfer:  %{time_pretransfer}\n
      time_redirect:  %{time_redirect}\n
 time_starttransfer:  %{time_starttransfer}\n
                    ----------\n
         time_total:  %{time_total}\n
EOF

Chiama allo stesso modo dell'alias:

curltime wordpress.org

Crea un collegamento di Windows (aka file BAT)

Inserisci questo comando in CURLTIME.BAT (nella stessa cartella di curl.exe)

curl -w "@%~dp0curl-format.txt" -o NUL -s %*

Quindi puoi semplicemente chiamare ...

curltime wordpress.org

Ecco la risposta:

curl -X POST -d @file server:port -w %{time_connect}:%{time_starttransfer}:%{time_total}

Tutte le variabili utilizzate con -wpossono essere trovate in man curl.

Opzione 1. Per misurare total time:

curl -o /dev/null -s -w 'Total: %{time_total}s\n'  https://www.google.com

Uscita campione:

Opzione 2. Per ottenere time to establish connection, TTFB: time to first bytee total time:

curl -o /dev/null -s -w 'Establish Connection: %{time_connect}s\nTTFB: %{time_starttransfer}s\nTotal: %{time_total}s\n'  https://www.google.com

Uscita campione:

Rif: Ottieni tempi di risposta con l'arricciatura

Una scorciatoia che puoi aggiungere al tuo .bashrc ecc., Sulla base di altre risposte qui:

function perf {
  curl -o /dev/null -s -w "%{time_connect} + %{time_starttransfer} = %{time_total}\n" "$1"
}

Uso:

> perf stackoverflow.com
0.521 + 0.686 = 1.290

Quanto segue è ispirato dalla risposta di Simon. È autonomo (non richiede un file di formato separato), il che lo rende perfetto per l'inclusione .bashrc.

curl_time() {
    curl -so /dev/null -w "\
   namelookup:  %{time_namelookup}s\n\
      connect:  %{time_connect}s\n\
   appconnect:  %{time_appconnect}s\n\
  pretransfer:  %{time_pretransfer}s\n\
     redirect:  %{time_redirect}s\n\
starttransfer:  %{time_starttransfer}s\n\
-------------------------\n\
        total:  %{time_total}s\n" "$@"
}

Inoltre, dovrebbe funzionare con tutti gli argomenti che curlnormalmente prendono, poiché i "$@"giusti li superano. Ad esempio, puoi fare:

curl_time -X POST -H "Content-Type: application/json" -d '{"key": "val"}' https://postman-echo.com/post

Produzione:

   namelookup:  0,125000s
      connect:  0,250000s
   appconnect:  0,609000s
  pretransfer:  0,609000s
     redirect:  0,000000s
starttransfer:  0,719000s
-------------------------
        total:  0,719000s

Se vuoi analizzare o riassumere la latenza puoi provare la panca apache:

ab -n [number of samples] [url]

Per esempio:

ab -n 100 http://www.google.com/

Mostrerà:

This is ApacheBench, Version 2.3 <$Revision: 1757674 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking www.google.com (be patient).....done


Server Software:        gws
Server Hostname:        www.google.com
Server Port:            80

Document Path:          /
Document Length:        12419 bytes

Concurrency Level:      1
Time taken for tests:   10.700 seconds
Complete requests:      100
Failed requests:        97
   (Connect: 0, Receive: 0, Length: 97, Exceptions: 0)
Total transferred:      1331107 bytes
HTML transferred:       1268293 bytes
Requests per second:    9.35 [#/sec] (mean)
Time per request:       107.004 [ms] (mean)
Time per request:       107.004 [ms] (mean, across all concurrent requests)
Transfer rate:          121.48 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:       20   22   0.8     22      26
Processing:    59   85 108.7     68     911
Waiting:       59   85 108.7     67     910
Total:         80  107 108.8     90     932

Percentage of the requests served within a certain time (ms)
  50%     90
  66%     91
  75%     93
  80%     95
  90%    105
  95%    111
  98%    773
  99%    932
 100%    932 (longest request)

Un altro modo è configurare in ~/.curlrcquesto modo

-w "\n\n==== cURL measurements stats ====\ntotal: %{time_total} seconds \nsize: %{size_download} bytes \ndnslookup: %{time_namelookup} seconds \nconnect: %{time_connect} seconds \nappconnect: %{time_appconnect} seconds \nredirect: %{time_redirect} seconds \npretransfer: %{time_pretransfer} seconds \nstarttransfer: %{time_starttransfer} seconds \ndownloadspeed: %{speed_download} byte/sec \nuploadspeed: %{speed_upload} byte/sec \n\n"

Quindi l'output di curlè

❯❯ curl -I https://google.com
HTTP/2 301
location: https://www.google.com/
content-type: text/html; charset=UTF-8
date: Mon, 04 Mar 2019 08:02:43 GMT
expires: Wed, 03 Apr 2019 08:02:43 GMT
cache-control: public, max-age=2592000
server: gws
content-length: 220
x-xss-protection: 1; mode=block
x-frame-options: SAMEORIGIN
alt-svc: quic=":443"; ma=2592000; v="44,43,39"



==== cURL measurements stats ====
total: 0.211117 seconds
size: 0 bytes
dnslookup: 0.067179 seconds
connect: 0.098817 seconds
appconnect: 0.176232 seconds
redirect: 0.000000 seconds
pretransfer: 0.176438 seconds
starttransfer: 0.209634 seconds
downloadspeed: 0.000 byte/sec
uploadspeed: 0.000 byte/sec

Hey è meglio di Apache Bench, ha meno problemi con SSL

./hey https://google.com -more
Summary:
  Total:    3.0960 secs
  Slowest:  1.6052 secs
  Fastest:  0.4063 secs
  Average:  0.6773 secs
  Requests/sec: 64.5992

Response time histogram:
  0.406 [1] |
  0.526 [142]   |∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
  0.646 [1] |
  0.766 [6] |∎∎
  0.886 [0] |
  1.006 [0] |
  1.126 [0] |
  1.246 [12]    |∎∎∎
  1.365 [32]    |∎∎∎∎∎∎∎∎∎
  1.485 [5] |∎
  1.605 [1] |

Latency distribution:
  10% in 0.4265 secs
  25% in 0.4505 secs
  50% in 0.4838 secs
  75% in 1.2181 secs
  90% in 1.2869 secs
  95% in 1.3384 secs
  99% in 1.4085 secs

Details (average, fastest, slowest):
  DNS+dialup:    0.1150 secs, 0.0000 secs, 0.4849 secs
  DNS-lookup:    0.0032 secs, 0.0000 secs, 0.0319 secs
  req write:     0.0001 secs, 0.0000 secs, 0.0007 secs
  resp wait:     0.2068 secs, 0.1690 secs, 0.4906 secs
  resp read:     0.0117 secs, 0.0011 secs, 0.2375 secs

Status code distribution:
  [200] 200 responses

Riferimenti

• https://github.com/rakyll/hey

Un'altra opzione che è forse la più semplice in termini di riga di comando è l'aggiunta --trace-timedell'opzione integrata :

curl -X POST -d @file server:port --trace-time

Anche se tecnicamente non genera i tempi dei vari passi come richiesto dall'OP, mostra i timestamp per tutti i passi della richiesta come mostrato di seguito. Usando questo, puoi (abbastanza facilmente) calcolare quanto tempo ha impiegato ogni passaggio.

$ curl https://www.google.com --trace-time -v -o /dev/null
13:29:11.148734 * Rebuilt URL to: https://www.google.com/
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     013:29:11.149958 *   Trying 172.217.20.36...
13:29:11.149993 * TCP_NODELAY set
13:29:11.163177 * Connected to www.google.com (172.217.20.36) port 443 (#0)
13:29:11.164768 * ALPN, offering h2
13:29:11.164804 * ALPN, offering http/1.1
13:29:11.164833 * successfully set certificate verify locations:
13:29:11.164863 *   CAfile: none
  CApath: /etc/ssl/certs
13:29:11.165046 } [5 bytes data]
13:29:11.165099 * (304) (OUT), TLS handshake, Client hello (1):
13:29:11.165128 } [512 bytes data]
13:29:11.189518 * (304) (IN), TLS handshake, Server hello (2):
13:29:11.189537 { [100 bytes data]
13:29:11.189628 * TLSv1.2 (IN), TLS handshake, Certificate (11):
13:29:11.189658 { [2104 bytes data]
13:29:11.190243 * TLSv1.2 (IN), TLS handshake, Server key exchange (12):
13:29:11.190277 { [115 bytes data]
13:29:11.190507 * TLSv1.2 (IN), TLS handshake, Server finished (14):
13:29:11.190539 { [4 bytes data]
13:29:11.190770 * TLSv1.2 (OUT), TLS handshake, Client key exchange (16):
13:29:11.190797 } [37 bytes data]
13:29:11.190890 * TLSv1.2 (OUT), TLS change cipher, Client hello (1):
13:29:11.190915 } [1 bytes data]
13:29:11.191023 * TLSv1.2 (OUT), TLS handshake, Finished (20):
13:29:11.191053 } [16 bytes data]
13:29:11.204324 * TLSv1.2 (IN), TLS handshake, Finished (20):
13:29:11.204358 { [16 bytes data]
13:29:11.204417 * SSL connection using TLSv1.2 / ECDHE-ECDSA-CHACHA20-POLY1305
13:29:11.204451 * ALPN, server accepted to use h2
13:29:11.204483 * Server certificate:
13:29:11.204520 *  subject: C=US; ST=California; L=Mountain View; O=Google LLC; CN=www.google.com
13:29:11.204555 *  start date: Oct  2 07:29:00 2018 GMT
13:29:11.204585 *  expire date: Dec 25 07:29:00 2018 GMT
13:29:11.204623 *  subjectAltName: host "www.google.com" matched cert's "www.google.com"
13:29:11.204663 *  issuer: C=US; O=Google Trust Services; CN=Google Internet Authority G3
13:29:11.204701 *  SSL certificate verify ok.
13:29:11.204754 * Using HTTP2, server supports multi-use
13:29:11.204795 * Connection state changed (HTTP/2 confirmed)
13:29:11.204840 * Copying HTTP/2 data in stream buffer to connection buffer after upgrade: len=0
13:29:11.204881 } [5 bytes data]
13:29:11.204983 * Using Stream ID: 1 (easy handle 0x55846ef24520)
13:29:11.205034 } [5 bytes data]
13:29:11.205104 > GET / HTTP/2
13:29:11.205104 > Host: www.google.com
13:29:11.205104 > User-Agent: curl/7.61.0
13:29:11.205104 > Accept: */*
13:29:11.205104 > 
13:29:11.218116 { [5 bytes data]
13:29:11.218173 * Connection state changed (MAX_CONCURRENT_STREAMS == 100)!
13:29:11.218211 } [5 bytes data]
13:29:11.251936 < HTTP/2 200 
13:29:11.251962 < date: Fri, 19 Oct 2018 10:29:11 GMT
13:29:11.251998 < expires: -1
13:29:11.252046 < cache-control: private, max-age=0
13:29:11.252085 < content-type: text/html; charset=ISO-8859-1
13:29:11.252119 < p3p: CP="This is not a P3P policy! See g.co/p3phelp for more info."
13:29:11.252160 < server: gws
13:29:11.252198 < x-xss-protection: 1; mode=block
13:29:11.252228 < x-frame-options: SAMEORIGIN
13:29:11.252262 < set-cookie: 1P_JAR=2018-10-19-10; expires=Sun, 18-Nov-2018 10:29:11 GMT; path=/; domain=.google.com
13:29:11.252297 < set-cookie: NID=141=pzXxp1jrJmLwFVl9bLMPFdGCtG8ySQKxB2rlDWgerrKJeXxfdmB1HhJ1UXzX-OaFQcnR1A9LKYxi__PWMigjMBQHmI3xkU53LI_TsYRbkMNJNdxs-caQQ7fEcDGE694S; expires=Sat, 20-Apr-2019 10:29:11 GMT; path=/; domain=.google.com; HttpOnly
13:29:11.252336 < alt-svc: quic=":443"; ma=2592000; v="44,43,39,35"
13:29:11.252368 < accept-ranges: none
13:29:11.252408 < vary: Accept-Encoding
13:29:11.252438 < 
13:29:11.252473 { [5 bytes data]
100 12215    0 12215    0     0   112k      0 --:--:-- --:--:-- --:--:--  112k
13:29:11.255674 * Connection #0 to host www.google.com left intact

curl -v --trace-time Questo deve essere fatto in modalità dettagliata

Ho creato un formattatore amichevole per annusare le richieste di arricciatura per aiutare con il debug (vedi commenti per l'uso). Contiene tutti i parametri di output noti che è possibile scrivere in un formato di facile lettura.

https://gist.github.com/manifestinteractive/ce8dec10dcb4725b8513

ecco la stringa che puoi usare con -w, contiene tutte le opzioni che curl -wsupportano.

{"contentType":"%{content_type}","filenameEffective":"%{filename_effective}","ftpEntryPath":"%{ftp_entry_path}","httpCode":"%{http_code}","httpConnect":"%{http_connect}","httpVersion":"%{http_version}","localIp":"%{local_ip}","localPort":"%{local_port}","numConnects":"%{num_connects}","numRedirects":"%{num_redirects}","proxySslVerifyResult":"%{proxy_ssl_verify_result}","redirectUrl":"%{redirect_url}","remoteIp":"%{remote_ip}","remotePort":"%{remote_port}","scheme":"%{scheme}","size":{"download":"%{size_download}","header":"%{size_header}","request":"%{size_request}","upload":"%{size_upload}"},"speed":{"download":"%{speed_download}","upload":"%{speed_upload}"},"sslVerifyResult":"%{ssl_verify_result}","time":{"appconnect":"%{time_appconnect}","connect":"%{time_connect}","namelookup":"%{time_namelookup}","pretransfer":"%{time_pretransfer}","redirect":"%{time_redirect}","starttransfer":"%{time_starttransfer}","total":"%{time_total}"},"urlEffective":"%{url_effective}"}

genera JSON.

Ecco un Bash one-liner per colpire ripetutamente lo stesso server:

for i in {1..1000}; do curl -s -o /dev/null -w "%{time_total}\n" http://server/get_things; done

Questa è una versione modificata della risposta di Simons che rende l'output multi-linea una singola riga. Presenta anche il timestamp corrente, quindi è più facile seguire ogni riga di output.

$ cat time-format.txt
time_namelookup:%{time_namelookup} time_connect:%{time_connect} time_appconnect:%{time_appconnect} time_pretransfer:%{time_pretransfer} time_redirect:%{time_redirect} time_starttransfer:%{time_starttransfer} time_total:%{time_total}\n

$ while [ 1 ];do echo -n "$(date) - " ; curl -w @curl-format.txt -o /dev/null -s https://myapp.mydom.com/v1/endpt-http; sleep 1; done | grep -v time_total:0

Mon Dec 16 17:51:47 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.172 time_pretransfer:0.172 time_redirect:0.000 time_starttransfer:1.666 time_total:1.666
Mon Dec 16 17:51:50 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.175 time_pretransfer:0.175 time_redirect:0.000 time_starttransfer:3.794 time_total:3.795
Mon Dec 16 17:51:55 UTC 2019 - time_namelookup:0.004 time_connect:0.017 time_appconnect:0.175 time_pretransfer:0.175 time_redirect:0.000 time_starttransfer:1.971 time_total:1.971
Mon Dec 16 17:51:58 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.173 time_pretransfer:0.173 time_redirect:0.000 time_starttransfer:1.161 time_total:1.161
Mon Dec 16 17:52:00 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.166 time_pretransfer:0.167 time_redirect:0.000 time_starttransfer:1.434 time_total:1.434
Mon Dec 16 17:52:02 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.177 time_pretransfer:0.177 time_redirect:0.000 time_starttransfer:5.119 time_total:5.119
Mon Dec 16 17:52:08 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.172 time_pretransfer:0.172 time_redirect:0.000 time_starttransfer:30.185 time_total:30.185
Mon Dec 16 17:52:39 UTC 2019 - time_namelookup:0.004 time_connect:0.014 time_appconnect:0.164 time_pretransfer:0.164 time_redirect:0.000 time_starttransfer:30.175 time_total:30.176
Mon Dec 16 17:54:28 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:3.191 time_pretransfer:3.191 time_redirect:0.000 time_starttransfer:3.212 time_total:3.212
Mon Dec 16 17:56:08 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:1.184 time_pretransfer:1.184 time_redirect:0.000 time_starttransfer:1.215 time_total:1.215
Mon Dec 16 18:00:24 UTC 2019 - time_namelookup:0.004 time_connect:0.015 time_appconnect:0.181 time_pretransfer:0.181 time_redirect:0.000 time_starttransfer:1.267 time_total:1.267

Ho usato quanto sopra per ottenere risposte lente sull'endpoint sopra.