smtp.test61 at gmail.com
Mon Aug 31 20:06:58 UTC 2009
Im testing AFR replication and noticed that a starnge thing probebly a
bug in gluster 2 , say there are 2 servers in AFR mode , 1 of them
then few files get modified on the remaining node , none of the
changes propergate when the otehr node comes back .
Then i noticed this on gluster faq "understaning AFR"
In other words, data replication only happens on those servers which
were alive when the file was opened. This is a rather tricky issue but
we hope to fix it very soon.
At first i thought its mispeled but then made some tests and all
confirmed this issue .
Current me if im wrong but this is not a "tricky issue" its making the
all AFR replication useless in all ways a can think of .
On production you can expect reboots from time to time also you can
expect file modifications happen non stop so what good is a replcation
that doesnt take care of those files that where changed .
New files can be created using ls -lR or other methods wich is not
very nice nut fine seens it works , but the issue with modified files
is realy making the AFR consept unussable .
Anyone can fix my way of thinking regarding this ? any work arround
for this issue .
I kind of like th gluster consept .
More information about the Gluster-users