Linux: Outils de monitoring/performances

Processeur

Top

Outil de base, top donne les performances globales, ou par coeur. Commandes intéressantes:

• 1 : bascule stats globales / par coeur.

Options intéressantes:

• -b -n 1 : cliché one-shot de tous les process.

Réseau

Ifop

iftop donne des stats sur une ou toutes les interfaces:

• sudo iftop -i ens0

nstat / rtacct

      Usage: nstat [ -h?vVzrnasd:t: ] [ PATTERN [ PATTERN ] ]
      Usage: rtacct [ -h?vVzrnasd:t: ] [ ListOfRealms ]

tracer les messages reseau sur un PID avec strace

Par exemple, avec l'option -f pour suivre la trace sur les process fils du process en trace:

 $ strace -v -f -e trace=network -p 4300
 strace: Process 4300 attached                                                                                                                                    
 socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT) = 3                                                                                                                  
 sendmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(2)=[{"\35\0\0\0d\4\1\0\0\0\0\0\0\0\0\0", 16}, {"ping 8.8.8.8\0", 13}], msg_controllen=0, msg_flags=0}, 0) = 29                                                                                                                                   
 strace: Process 23396 attached                                                                                                                                   
 [pid 23396] socket(AF_INET, SOCK_DGRAM, IPPROTO_ICMP) = -1 EACCES (Permission denied)                                                                            
 [pid 23396] socket(AF_INET, SOCK_RAW, IPPROTO_ICMP) = 3                                                                                                          
 [pid 23396] socket(AF_INET, SOCK_DGRAM, IPPROTO_IP) = 4                                                                                                          
 [pid 23396] connect(4, {sa_family=AF_INET, sin_port=htons(1025), sin_addr=inet_addr("8.8.8.8")}, 16) = 0                                                         
 [pid 23396] getsockname(4, {sa_family=AF_INET, sin_port=htons(58430), sin_addr=inet_addr("192.168.1.16")}, [16]) = 0                                             
 [pid 23396] setsockopt(3, SOL_RAW, ICMP_FILTER, ~(1<<ICMP_ECHOREPLY|1<<ICMP_DEST_UNREACH|1<<ICMP_SOURCE_QUENCH|1<<ICMP_REDIRECT|1<<ICMP_TIME_EXCEEDED|1<<ICMP_PARAMETERPROB), 4) = 0                                                                                                                                              
 [pid 23396] setsockopt(3, SOL_IP, IP_RECVERR, [1], 4) = 0
 [pid 23396] setsockopt(3, SOL_SOCKET, SO_SNDBUF, [324], 4) = 0
 [pid 23396] setsockopt(3, SOL_SOCKET, SO_RCVBUF, [65536], 4) = 0
 [pid 23396] getsockopt(3, SOL_SOCKET, SO_RCVBUF, [131072], [4]) = 0
 [pid 23396] setsockopt(3, SOL_SOCKET, SO_TIMESTAMP, [1], 4) = 0
 [pid 23396] setsockopt(3, SOL_SOCKET, SO_SNDTIMEO, "\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0", 16) = 0
 [pid 23396] setsockopt(3, SOL_SOCKET, SO_RCVTIMEO, "\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0", 16) = 0
 [pid 23396] sendto(3, "\10\0\33\21[d\0\1\237\10wZ\0\0\0\0\242S\n\0\0\0\0\0\20\21\22\23\24\25\26\27"..., 64, 0, {sa_family=AF_INET, sin_port=htons(0), sin_addr=inet_addr("8.8.8.8")}, 16) = 64
 [pid 23396] recvmsg(3, {msg_name(16)={sa_family=AF_INET, sin_port=htons(0), sin_addr=inet_addr("8.8.8.8")}, msg_iov(1)=[{"E\0\0T\321\360\0\0;\1\333\360\10\10\10\10\300\250\1\20\0\0#\21[d\0\1\237\10wZ"..., 192}], msg_controllen=32, [{cmsg_len=32, cmsg_level=SOL_SOCKET, cmsg_type=0x1d /* SCM_??? */}], msg_flags=0}, 0) = 84
 [pid 23396] --- SIGINT {si_signo=SIGINT, si_code=SI_KERNEL} ---
[pid 23396] +++ exited with 0 +++
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=23396, si_uid=1000, si_status=0, si_utime=0, si_stime=0} ---

outils eBPF (kernel > 4.2)

Voir ces mécanismes : https://opensource.com/article/17/11/bccbpf-performance