Remove Hundreds of Thousands of Files, take 3
After blogged about using find to remove files, I realised that there are thousands of error in execve system calls. By simulating the same scenario with just 2 files, I understand that the search PATH is the culprit.
[pid 29481] execve("/usr/lib/lightdm/lightdm/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/usr/local/sbin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/usr/local/bin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/usr/sbin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/usr/bin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/sbin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = -1 ENOENT (No such file or directory)
[pid 29481] execve("/bin/rm", ["rm", "-f", "./somefiles-a"], [/* 50 vars */]) = 0
The amount of system calls is reduced (especially errors) if I specify the correct full path of 'rm'.
$ touch somefiles-{a..z}{a..z}{a..z}
$strace -cf find . -name "somefiles-*" -exec /bin/rm -f {} \;
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
96.59 36.418146 2072 17576 waitpid
2.50 0.940861 54 17576 clone
0.72 0.271665 15 17576 unlinkat
0.07 0.025220 0 105467 close
0.03 0.012440 541 23 getdents64
0.03 0.011857 1 17577 fstatat64
0.02 0.009425 1 17576 17576 _llseek
0.01 0.005058 0 52737 open
0.01 0.004589 0 140626 mmap2
0.01 0.003785 0 17577 ioctl
0.00 0.001053 0 52757 brk
0.00 0.001028 0 52735 fstat64
0.00 0.000388 0 17577 munmap
0.00 0.000245 0 70311 mprotect
0.00 0.000000 0 17580 read
0.00 0.000000 0 17577 execve
0.00 0.000000 0 52734 52734 access
0.00 0.000000 0 1 gettimeofday
0.00 0.000000 0 2 uname
0.00 0.000000 0 17581 fchdir
0.00 0.000000 0 3 rt_sigaction
0.00 0.000000 0 1 rt_sigprocmask
0.00 0.000000 0 2 getrlimit
0.00 0.000000 0 2 1 futex
0.00 0.000000 0 17577 set_thread_area
0.00 0.000000 0 1 set_tid_address
0.00 0.000000 0 1 openat
0.00 0.000000 0 17577 set_robust_list
------ ----------- ----------- --------- --------- ----------------
100.00 37.705760 738330 70311 total
Another way to further reduce the amount of system calls as well as run time is to take advantage of 'rm' ability to take more than one file as argument. Getting 'find' output to pipe to 'xargs -L 10 /bin/rm -f', we are able to ask 'rm' to remove 10 files at a time. You can see the mass reduction in system calls and run time.
$ cat rm.sh
#! /bin/bash
find . -name "somefiles-*" | xargs -L 10 /bin/rm -f
$ touch somefiles-{a..z}{a..z}{a..z}
$ strace -cf ./rm.sh
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
98.63 9.880623 5611 1761 1 waitpid
0.98 0.097799 6 17576 unlinkat
0.18 0.018443 10 1760 clone
0.12 0.011728 510 23 getdents64
0.08 0.008373 0 17577 fstatat64
0.00 0.000331 0 7058 4 open
0.00 0.000312 0 1842 read
0.00 0.000289 0 14108 mmap2
0.00 0.000041 0 1761 1760 ioctl
0.00 0.000039 0 12342 2 close
0.00 0.000000 0 69 write
0.00 0.000000 0 1761 execve
0.00 0.000000 0 1 time
0.00 0.000000 0 1 getpid
0.00 0.000000 0 5296 5288 access
0.00 0.000000 0 1 pipe
0.00 0.000000 0 5323 brk
0.00 0.000000 0 3 dup2
0.00 0.000000 0 1 getppid
0.00 0.000000 0 1 getpgrp
0.00 0.000000 0 2 gettimeofday
0.00 0.000000 0 1765 munmap
0.00 0.000000 0 1 sigreturn
0.00 0.000000 0 3 uname
0.00 0.000000 0 7049 mprotect
0.00 0.000000 0 5 fchdir
0.00 0.000000 0 1761 _llseek
0.00 0.000000 0 25 rt_sigaction
0.00 0.000000 0 21 rt_sigprocmask
0.00 0.000000 0 5 getrlimit
0.00 0.000000 0 24 8 stat64
0.00 0.000000 0 5295 fstat64
0.00 0.000000 0 9 getuid32
0.00 0.000000 0 9 getgid32
0.00 0.000000 0 9 geteuid32
0.00 0.000000 0 9 getegid32
0.00 0.000000 0 3 1 fcntl64
0.00 0.000000 0 2 1 futex
0.00 0.000000 0 1761 set_thread_area
0.00 0.000000 0 1 set_tid_address
0.00 0.000000 0 1 openat
0.00 0.000000 0 1 set_robust_list
------ ----------- ----------- --------- --------- ----------------
100.00 10.017978 106026 7065 total
posted by chihungchan at 4:25 AM
0 Comments:
Post a Comment
<< Home