[Gluster-users] replicate issue
zhuoyin at gmail.com
Fri Sep 25 06:11:34 UTC 2009
I met an issue when dealing with replicate translator.
I have 48 x 1TB disks, they are lied on several machines. I configured
16 volumes, each volumes is based on replicate translator and
contained 3 disks (copies) from 3 different machines. So the efficient
storage space is 16T. each machine are using the exactly the same
client/server configuration file and all the machines can access to
the whole space.
I'm testing the glusterfs by reading files, and at the same time, I
randomly bring down two machines' NICs to simulate machine's shut
down. since my configuration ensured 3 copies of the same data are all
on the different machines, so fail two machines will not affect the
But in my test, it fails, I read a file via machine A, and this file
has 3 copies in 3 different machines (machine B, C, D), I failed the
machine B and C's NIC, and machine D is still on-line and the copy of
file is still there. when I doing `stat foo.txt` on machine A, it
reports can't find the file. I think it's not working as expect on
The tricky things is, machine A, D are using exactly the same
configuration as I mentioned before, so both can access the 16T space.
And I tried to do `stat foo.txt` on machine D which has the last copy
of foo.txt. it finally found the file, and I do `stat foo.txt` on
machine A again, it also found this file.
More information about the Gluster-users