[Gluster-users] The troubles with invisible file but there is in brick directory

kenji kondo kkay.jp at gmail.com
Wed Jul 18 03:15:33 UTC 2012


Dear All,

I tested the GlusterFS 3.2.6 on CentOS 5.8. It was good performance.
But some troubles are occurring on GlusterFS 3.2.7 and CentOS 6.2...

I can write files and make directories. But, the trouble is when I use
'ls' command,
some files become invisible and often ls command could not return.

To show it, I tried to make the trouble on one host.
It likes as following:

---Create volume
gluster> volume info
Volume Name: v1
Type: Distribute
Status: Started
Number of Bricks: 3
Transport-type: tcp
Bricks:
Brick1: host14:/mnt/a1
Brick2: host14:/mnt/a2
Brick3: host14:/mnt/a3

---Mount above volume
[kk at host14 /mnt]$ sudo mount -t glusterfs host14:v1 /mnt/v1

---Create a file in the mounted directory but it becomes invisible for
ls command, but there is it in brick
[kk at host14 /mnt]$ cd v1
[kk at host14 v1]$ ls
[kk at host14 v1]$ touch a
[kk at host14 v1]$ ls
[kk at host14 v1]$ ls ../a?
../a1:

../a2:
a

../a3:
[kk at host14 v1]$ rm a
[kk at host14 v1]$ ls ../a?
../a1:

../a2:

../a3:
---Create some files but it becomes to not return for ls command
[kk at host14 v1]$ touch a b c d e
[kk at host14 v1]$ ls ../a?
../a1:
d

../a2:
a c

../a3:
b e
[kk at host14 v1]$ ls

[Here, I need to kill the process from other console.]

Killed
[kk at host14 v1]$


For this no return, I confirmed that a "glusterfs" and three
"glusterfsd" processes have high CPU load after 'ls' command run.
If the volume has two bricks, this trouble doesn't occur.
I can understand that this situation is too strange. So I am thinking
there is something simple mistake for setting on my system.
I checked log files, but I could not find the clue to solve.
If anybody knows about this, could you please tell me?

Best regards,
K. Kondo



More information about the Gluster-users mailing list