toolkits / nux Goto Github PK

View Code? Open in Web Editor NEW

237.0 22.0 83.0 38 KB

*nux metric collector

Go 100.00%

nux's Issues

fuse type ignore

my disk status:
df -Th
Filesystem Type Size Used Avail Use% Mounted on
ceph-fuse fuse.ceph-fuse 20T 20T 626G 97% /data

https://github.com/toolkits/nux/blob/master/dfstat_linux.go
if strings.HasPrefix(fsVfstype, "fuse") {
continue
}

why ignore "fuse" prefix?

why keep /dev/xxx device with shorter fsFile (remove mount binds)?

nux/dfstat_linux.go

Line 63 in 4ddb000

// keep /dev/xxx device with shorter fsFile (remove mount binds)


				if ret[idx][0] == fsSpec {
					deviceFound = true
					if len(fsFile) < len(ret[idx][1]) {
						ret[idx][1] = fsFile
					}
					break
				}

cmdline 如果被进程重写，现有逻辑将无法忽略空格。

nux/proc.go

Line 67 in debb382

if cmdlineBytes[j] != 0 {

有一些进程，例如nginx会重写cmdline：

[root@vm-vm114 falcon-agent]# ps aux|grep nginx
root     14811  0.0  0.0 108964  1884 ?        Ss   19:18   0:00 nginx: master process /usr/sbin/nginx -c /etc/nginx/nginx.conf
nginx    14812  0.0  0.1 109388  2736 ?        S    19:18   0:00 nginx: worker process
nginx    14813  0.0  0.1 109388  2656 ?        S    19:18   0:00 nginx: worker process
root     15182  0.0  0.0 103336   864 pts/2    S+   19:57   0:00 grep nginx
[root@vm-vm114 falcon-agent]#
[root@vm-vm114 falcon-agent]# xxd /proc/14811/cmdline
0000000: 6e67 696e 783a 206d 6173 7465 7220 7072  nginx: master pr
0000010: 6f63 6573 7320 2f75 7372 2f73 6269 6e2f  ocess /usr/sbin/
0000020: 6e67 696e 7820 2d63 202f 6574 632f 6e67  nginx -c /etc/ng
0000030: 696e 782f 6e67 696e 782e 636f 6e66       inx/nginx.conf

可以看到，原本看起来应该是0x00的地方，却用0x20(空格字符)，falcon-agent在使用这个库获取到的进程含有空格，会导致cmdline无法匹配上。

Need more test like goprocinfo

Please refer to https://github.com/c9s/goprocinfo

Kernel 4.18+ appends four more fields putting the total at 18

nux/iostat.go

Line 72 in d5bf087

if size != 14 {

Kernel 4.18+ appends four more fields for discard tracking putting the total at 18:
15 - discards completed successfully
16 - discards merged
17 - sectors discarded
18 - time spent discarding

https://www.kernel.org/doc/Documentation/ABI/testing/procfs-diskstats

crash on Fedora

ss_s.go

arr := strings.Split(content, ", ")
                        for _, val := range arr {
                                fields := strings.Fields(val)
                                if fields[0] == "timewait" {
                                        timewait_arr := strings.Split(fields[1], "/")
                                        m["timewait"], _ = strconv.ParseUint(timewait_arr[0], 10, 64)
                                        m["slabinfo.timewait"], _ = strconv.ParseUint(timewait_arr[1], 10, 64)
                                        continue
                                }
                                m[fields[0]], _ = strconv.ParseUint(fields[1], 10, 64)
                        }
                        return

ss -s on Fedora returns:

Total: 843
TCP:   89 (estab 56, closed 11, orphaned 0, timewait 2)

Transport Total     IP        IPv6
RAW	  1         0         1        
UDP	  8         5         3        
TCP	  78        48        30       
INET	  87        53        34       
FRAG	  0         0         0

while on CentOS, it returns:

Total: 525 (kernel 0)
TCP:   14 (estab 8, closed 0, orphaned 0, synrecv 0, timewait 0/0), ports 0

Transport Total     IP        IPv6
*	  0         -         -        
RAW	  0         0         0        
UDP	  7         4         3        
TCP	  14        11        3        
INET	  21        15        6        
FRAG	  0         0         0

So, it always crashes on Fedora with:

panic: runtime error: index out of range

Double count Guest Time in Total CPU Time

When running Virtual Machine(KVM) in host, I find that the value of host cpu usage is different between falcon agent and top cmd. Actually, falcon agent use nux package to calculate cpu usage.
https://github.com/open-falcon-archive/agent/blob/master/funcs/cpustat.go#L19

In Linux Kernel, User time is already include Guest Time. So when read cpu info from /proc/stat, we should not add guest time once more.

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/kernel/sched/cputime.c#n138

ignore file type: squashfs

Ubuntu 18.04 default running snapd process, which auto mount squashfs files.

http://squashfs.sourceforge.net/

Squashfs is a compressed read-only filesystem for Linux. Squashfs is intended for general read-only filesystem use, for archival use (i.e. in cases where a .tar.gz file may be used), and in constrained block device/memory systems (e.g. embedded systems) where low overhead is needed.

as squashfs is a read-only filesystem, mounting a squashfs file likes mounting a read-pnly cd/dvd iso file. may be we just ignore this filetype.

#df -h
Filesystem                         Size  Used Avail Use% Mounted on
udev                               962M     0  962M   0% /dev
tmpfs                              199M  1.1M  198M   1% /run
/dev/mapper/ubuntu--vg-ubuntu--lv  118G   11G  108G  10% /
tmpfs                              992M   12K  992M   1% /dev/shm
tmpfs                              5.0M     0  5.0M   0% /run/lock
tmpfs                              992M     0  992M   0% /sys/fs/cgroup
/dev/loop0                          90M   90M     0 100% /snap/core/6130
/dev/loop1                          90M   90M     0 100% /snap/core/6673
/dev/loop2                          89M   89M     0 100% /snap/core/7396
/dev/xvda2                         976M  143M  767M  16% /boot
tmpfs                              199M     0  199M   0% /run/user/47483
#df -i
Filesystem                          Inodes  IUsed    IFree IUse% Mounted on
udev                                246153    430   245723    1% /dev
tmpfs                               253889    670   253219    1% /run
/dev/mapper/ubuntu--vg-ubuntu--lv 61865984 181453 61684531    1% /
tmpfs                               253889      4   253885    1% /dev/shm
tmpfs                               253889      3   253886    1% /run/lock
tmpfs                               253889     18   253871    1% /sys/fs/cgroup
/dev/loop0                           12810  12810        0  100% /snap/core/6130
/dev/loop1                           12819  12819        0  100% /snap/core/6673
/dev/loop2                           12823  12823        0  100% /snap/core/7396
/dev/xvda2                           65536    309    65227    1% /boot
tmpfs                               253889     10   253879    1% /run/user/47483

ignore file type: iso9660

mount an iso file, the filesystem info show below.
as iso files are read-only, mayby we can just ignore them.

mount -o loop ubuntu-18.04.2-server-amd64.iso  /mnt/

df -h
/dev/loop0                  883M  883M     0 100% /mnt

df -i
/dev/loop0                        0      0        0     - /mnt


fs_spec: /dev/loop0
fs_file: /mnt
fs_vfstype: iso9660

Unable to read the host file system while docker is running

When reading files, the file path can be prefixed to adapt to the file system of the host in docker.

dfstat 文件类型过滤

dfstat.go

centos：

proc /run/docker/netns/1fea574dae2e proc rw,nosuid,nodev,noexec,relatime 0 0

Ubuntu：

nsfs /run/docker/netns/b6fb8a43129f nsfs rw 0 0

Ubuntu 多了一个 nsfs 的类型
是否考虑加 FSTYPE_IGNORE 列表中

toolkits / nux Goto Github PK

nux's Issues

fuse type ignore

why keep /dev/xxx device with shorter fsFile (remove mount binds)?

cmdline 如果被进程重写，现有逻辑将无法忽略空格。

Need more test like goprocinfo

Kernel 4.18+ appends four more fields putting the total at 18

crash on Fedora

Double count Guest Time in Total CPU Time

ignore file type: squashfs

ignore file type: iso9660

Unable to read the host file system while docker is running

dfstat 文件类型过滤

dfstat.go

centos：

Ubuntu：

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent