Shulou(Shulou.com)06/01 Report--

How to monitor the real-time IO situation of Linux process, many novices are not very clear about this, in order to help you solve this problem, the following small series will explain in detail for everyone, there are people who need this to learn, I hope you can gain something.

Linux Kernel 2.6.20 and above kernel support process IO statistics, you can use tools like iotop to monitor each process on IO operations, just like using top to view the process memory, CPU and other occupation in real time. Linux kernel versions below 2.6.20 are not so lucky. I wrote a simple Python script to run Linux kernel.

< 2.6.20 下打印进程 IO 状况。 Kernel < 2.6.20 这个脚本的想法很简单，把 dmesg 的结果重定向到一个文件后再解析出来，每隔1秒钟打印一次进程 IO 读写的统计信息，执行这个脚本需要 root： 1. #!/usr/bin/python 2. # Monitoring per-process disk I/O activity 3. # written by http://www.vpsee.com 4. 5. import sys, os, time, signal, re 6. 7. class DiskIO: 8. def __init__(self, pname=None, pid=None, reads=0, writes=0): 9. self.pname = pname 10. self.pid = pid 11. self.reads = 0 12. self.writes = 0 13. 14. def main(): 15. argc = len(sys.argv) 16. if argc != 1: 17. print "usage: ./iotop" 18. sys.exit(0) 19. 20. if os.getuid() != 0: 21. print "must be run as root" 22. sys.exit(0) 23. 24. signal.signal(signal.SIGINT, signal_handler) 25. os.system('echo 1 >

/proc/sys/vm/block_dump')

26. print "TASK PID READ WRITE"

27. while True:

28. os.system('dmesg -c > /tmp/diskio.log')

29. l = []

30. f = open('/tmp/diskio.log', 'r')

31. line = f.readline()

32. while line:

33. m = re.match(\

34. '^(\S+)\((\d+)\): (READ|WRITE) block (\d+) on (\S+)', line)

35. if m != None:

36. if not l:

37. l.append(DiskIO(m.group(1), m.group(2)))

38. line = f.readline()

39. continue

40. found = False

41. for item in l:

42. if item.pid == m.group(2):

43. found = True

44. if m.group(3) == "READ":

45. item.reads = item.reads + 1

46. elif m.group(3) == "WRITE":

47. item.writes = item.writes + 1

48. if not found:

49. l.append(DiskIO(m.group(1), m.group(2)))

50. line = f.readline()

51. time.sleep(1)

52. for item in l:

53. print "%-10s s d d" % \

54. (item.pname, item.pid, item.reads, item.writes)

55.

56. def signal_handler(signal, frame):

57. os.system('echo 0 > /proc/sys/vm/block_dump')

58. sys.exit(0)

59.

60. if __name__=="__main__":

61. main()

62.

Kernel >= 2.6.20

If you want to use iotop to see process IO activity in real time, you need to download and upgrade the new kernel (version 2.6.20 or above). The TASK_DELAY_ACCT and TASK_IO_ACCOUNTING options need to be turned on when compiling a new kernel. After extracting the kernel, enter the configuration interface:

# tar jxvf linux-2.6.30.5.tar.bz2

# mv linux-2.6.30.5 /usr/src/

# cd /usr/src/linux-2.6.30.5

# make menuconfig

Select Kernel hacking -> Collect scheduler debugging info and Collect scheduler statistics to save the kernel and compile the kernel:

# make; make modules; make modules_install; make install

Modify the grub to make sure the new kernel starts correctly:

# vi /boot/grub/menu.lst

In addition to the new kernel, iotop requires Python 2.5 or above to run, so download and install the latest Python packages if your current Python is 2.4. Here is the source code compiled to install:

# tar jxvf Python-2.6.2.tar.bz2

# cd Python-2.6.2

# ./ configure

# make; make install

Don't forget to download the setuptools:

# mv setuptools-0.6c9-py2.6.egg.sh setuptools-0.6c9-py2.6.egg

# sh setuptools-0.6c9-py2.6.egg

Some netizens raised questions about the above script and asked why WRITE appeared 0. This is a good question. The author will explain it here. Let's first look at how we can monitor IO activity in real time across different processes.

block_dump

The Linux kernel provides a block_dump parameter to dump the WRITE/READ status to the log, which can be viewed by the dmesg command. The specific steps are:

# sysctl vm.block_dump=1

or

# echo 1 > /proc/sys/vm/block_dump

Then you can observe the IO activity status of each process through dmesg:

# dmesg -c

kjournald(542): WRITE block 222528 on dm-0

kjournald(542): WRITE block 222552 on dm-0

bash(18498): dirtied inode 5892488 (ld-linux-x86-64.so.2) on dm-0

bash(18498): dirtied inode 5892482 (ld-2.5.so) on dm-0

dmesg(18498): dirtied inode 11262038 (ld.so.cache) on dm-0

dmesg(18498): dirtied inode 5892496 (libc.so.6) on dm-0

dmesg(18498): dirtied inode 5892489 (libc-2.5.so) on dm-0

problem

A careful netizen raised the question: Why does WRITE block 0 appear? The author tracked it for a while and found that WRITE 0 did appear, such as:

# dmesg -c

...

pdflush(23123): WRITE block 0 on sdb1

pdflush(23123): WRITE block 16 on sdb1

pdflush(23123): WRITE block 104 on sdb1

pdflush(23123): WRITE block 40884480 on sdb1

...

answer

It turns out that we misunderstood the numbers contained in WRITE block 0, WRITE block 16, WRITE block 104. These numbers do not represent how many blocks were written, but which block was written. In order to find the truth, the author went to Linux 2.6.18 kernel code and found the answer in ll_rw_blk.c.:

$ vi linux-2.6.18/block/ll_rw_blk.c

1. void submit_bio(int rw, struct bio *bio)

2. {

3. int count = bio_sectors(bio);

4.

5. BIO_BUG_ON(! bio->bi_size);

6. BIO_BUG_ON(! bio->bi_io_vec);

7. bio->bi_rw |= rw;

8. if (rw & WRITE)

9. count_vm_events(PGPGOUT, count);

10. else

11. count_vm_events(PGPGIN, count);

12.

13. if (unlikely(block_dump)) {

14. char b[BDEVNAME_SIZE];

15. printk(KERN_DEBUG "%s(%d): %s block %Lu on %s\n",

16. current->comm, current->pid,

17. (rw & WRITE) ? "WRITE" : "READ",

18. (unsigned long long)bio->bi_sector,

19. bdevname(bio->bi_bdev,b));

20. }

21.

22. generic_make_request(bio);

23. }

It is obvious from the above code that WRITE block 0 on sdb1, where 0 is bio->bi_sector, which sector is written to, not how many blocks WRITE means. Also, if the block device is divided into multiple partitions, the bi_sector (sector number) is counted from the partition, for example, block 0 on sdb1 is the zeroth sector on sdb1.

Did reading the above help you? If you still want to have further understanding of related knowledge or read more related articles, please pay attention to the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.