mdadm multipath
/dev/xvda8 1583 1595 104391 83 Linux
/dev/xvda9 1596 1608 104391 83 Linux
/dev/xvda10 1609 1621 104391 83 Linux
멀티패스는 하드웨어 고장시 다른 경로로 데이터를 계속 주고 받을 수 있도록 하기 위함임.
/etc/mdadm.conf
DEVICE /dev/xvda[89],/dev/xvda10
ARRAY /dev/md1 devices=/dev/xvda8,/dev/xvda9,/dev/xvda10
[root@www ~]# mdadm -C /dev/md1 --level=multipath --raid-devices=3 /dev/xvda8 /dev/xvda9 /dev/xvda10
mdadm: array /dev/md1 started.
multipath 지정하면 raid0 으로 된다.
[root@www ~]# cat /proc/mdstat
Personalities : [raid0] [multipath]
md1 : active multipath xvda10[0] xvda9[1] xvda8[2]
104320 blocks [3/3] [UUU]
[root@www ~]# mdadm --detail /dev/md1
/dev/md1:
Version : 0.90
Creation Time : Wed Sep 14 16:03:47 2011
Raid Level : multipath
Array Size : 104320 (101.89 MiB 106.82 MB)
Raid Devices : 3
Total Devices : 3
Preferred Minor : 1
Persistence : Superblock is persistent
Update Time : Wed Sep 14 16:03:47 2011
State : clean
Active Devices : 3
Working Devices : 3
Failed Devices : 0
Spare Devices : 0
UUID : 6a4418e4:1171167a:645e0ce3:753772a6
Events : 0.3
Number Major Minor RaidDevice State
0 202 10 0 active sync /dev/xvda10
1 202 9 1 active sync /dev/xvda9
2 202 8 2 active sync /dev/xvda8
/dev/xvda8 이 문제가 있을 경우 어떻게 되는지 확인해 보자.
아래 옵션을 사용한다.
-f, --fail
-r, --remove
-a, --add
다른점은 공간이 하드 1개의 공간밖에 안된다는것이다.
멀티패스의 장점이 있지만, 3개의 공간 문제는 있다. 그러나 데이터 안정성을 위해서는
괜찮은 것 같다.
/dev/md1 99M 5.6M 89M 6% /data1
/dev/xvda8 을 fail로 만든다.
[root@www ~]# mdadm /dev/md1 -f /dev/xvda8
mdadm: set /dev/xvda8 faulty in /dev/md1
/data1 파일쓰기가 가능하다.
[root@www ~]# cd /data1
[root@www data1]# ls
lost+found
[root@www data1]# touch 1
[root@www data1]# ls -al
total 21
drwxr-xr-x 3 root root 1024 Sep 14 16:10 .
drwxr-xr-x 24 root root 4096 Sep 14 16:08 ..
-rw-r--r-- 1 root root 0 Sep 14 16:10 1
drwx------ 2 root root 12288 Sep 14 16:08 lost+found
문제있는 하드를 제거
[root@www data1]# mdadm /dev/md1 -r /dev/xvda8
mdadm: hot removed /dev/xvda8
다시 하드를 리스트에 추가
[root@www data1]# mdadm /dev/md1 -a /dev/xvda8
mdadm: re-added /dev/xvda8
시스템 메세지에 나타난 부분
Sep 14 16:03:47 www kernel: md1: WARNING: xvda9 appears to be on the same physical disk as xvda8. True
Sep 14 16:03:47 www kernel: protection against single-disk failure might be compromised.
Sep 14 16:03:47 www kernel: md: bind<xvda9>
Sep 14 16:03:47 www kernel: md1: WARNING: xvda10 appears to be on the same physical disk as xvda9. True
Sep 14 16:03:47 www kernel: protection against single-disk failure might be compromised.
Sep 14 16:03:47 www kernel: md: bind<xvda10>
Sep 14 16:03:47 www kernel: md: multipath personality registered for level -4
Sep 14 16:03:47 www kernel: multipath: array md1 active with 3 out of 3 IO paths
Sep 14 16:09:08 www kernel: kjournald starting. Commit interval 5 seconds
Sep 14 16:09:08 www kernel: EXT3 FS on md1, internal journal
Sep 14 16:09:08 www kernel: EXT3-fs: mounted filesystem with ordered data mode.
Sep 14 16:10:10 www kernel: multipath: IO failure on xvda8, disabling IO path.
Sep 14 16:10:10 www kernel: Operation continuing on 2 IO paths.
Sep 14 16:10:10 www kernel: MULTIPATH conf printout:
Sep 14 16:10:10 www kernel: --- wd:2 rd:3
Sep 14 16:10:10 www kernel: disk0, o:1, dev:xvda10
Sep 14 16:10:10 www kernel: disk1, o:1, dev:xvda9
Sep 14 16:10:10 www kernel: disk2, o:0, dev:xvda8
Sep 14 16:10:10 www kernel: MULTIPATH conf printout:
Sep 14 16:10:10 www kernel: --- wd:2 rd:3
Sep 14 16:10:10 www kernel: disk0, o:1, dev:xvda10
Sep 14 16:10:10 www kernel: disk1, o:1, dev:xvda9
Sep 14 16:11:19 www kernel: md: unbind<xvda8>
Sep 14 16:11:19 www kernel: md: export_rdev(xvda8)
Sep 14 16:13:28 www kernel: md1: WARNING: xvda8 appears to be on the same physical disk as xvda10. True
Sep 14 16:13:28 www kernel: protection against single-disk failure might be compromised.
Sep 14 16:13:28 www kernel: md: bind<xvda8>
|
댓글을 작성하시려면 로그인이 필요합니다.