[Gluster-users] gluster v3.0.0 log showing errors

IQnection Hosting Department hosting at iqnection.com
Sun Jan 17 23:20:06 UTC 2010


Recently we've put a gluster v3.0.0 file system into use for a group of  
apache2/tomcat6 servers.
Today one of those servers started hanging and we noticed a lot of lines  
like the following two in the log file:

1) W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 285653: GETATTR  
140556365266128 (fuse_loc_fill() failed)

2) E [afr.c:179:afr_read_child] mirror-0: invalid argument: inode

I don't think we're doing anything very strange with this setup. I'm at a  
loss as to why we are seeing this "hanging" behavior and these errors in  
the logs.

All machines are running v3.0.0 from the same source build.
We have 3 bricks using jfs for their backend storage setup to replicate to  
all three.
During testing we discovered that we needed to turn off write-behind to  
make things like CVS and xsl transformations more reliable in our setup.  
Write speed of course suffers but that is a compromise we've been willing  
to live with.
There are 7 clients using the file system for web files, java classes, and  
log files.
While this was going on there was an rsync of more images and PDF files  
copying from an old web server onto the new gluster file system using the  
client that was "hanging" as it's receiving side.

If you could shed any light on what might be causing our problems, it  
would be very helpful. Thanks!

Kind regards,

Chris Ely
Hosting Department
IQnection Internet Services, Inc.
23 Taylor Avenue
Doylestown, PA  18901
Ph. 1-866-400-4678/Fax. 215-489-1872

More information about the Gluster-users mailing list