[Gluster-users] slow directory listings, usage timeouts on glusterfs 3.0.4

[it-news (Josef Lahmer)]

Hi all!,

currently i'm "filling" glusterfs storage with data, while one client is
using it (serving a website too)...

client1:
rsync -av --delete /var/www/ /gFS/d/client1/ (6,7 GB with 60000 files)

client2:
rsync -av --delete /home/users/ root at client4:/gFS/d/ client2/ (7,3 gb
with 200000 files)

apache serveing a website from /gFS/d/client3/ directory

after 5 minutes rsync getting sync slowdowns and apache serving
errors/timeouts....
then i kill rsync processes, waiting 1 minute apache is working
correct....


what i don't understand:
every client is working in a separate directories, but ist still
affectet (slowdowns) 
=> could that be file locks??

...
is that a bug or a performance issue?
maybe special config for "many small files"?
xfs as server storage filesystem?
I've nothing found in glusterfs server & client logs, 
I've nothing found in system logs (var/log/messages)
any ideas?
...

regards from austria,
josy

details:
config:
glusterfs-volgen -c /etc/glusterfs/ --name glusterFSr1 --raid 1
gluster01:/glusterFS/r1 gluster02:/glusterFS/r1
glusterfs 3.0.4 built on Apr 14 2010 08:20:28

servers:
2xgluster server centos 5.5 (gluster01, gluster02), 2gb ram
store: 200 gb: /dev/xvdb on /glusterFS type xfs (rw,noatime)
2.6.18-194.3.1.el5xen

clients: 3*centos 5.5(client2-4), 1* debian (client1), 4gb ram
centos 5.5 (Linux host 2.6.18-194.3.1.el5xen #1 SMP x86_64 x86_64 x86_64
GNU/Linux)
debian 5.0.4 2.6.18.xs4.0.1.900.5799 #2 SMP

all systems and servers are in "virtual" environment (citrix xenserver)


gugler* News der Woche: Erfolgreiches Öko-Audit! Das g* Umweltmanagementsystem erfüllt die höchsten europäischen EMAS-Standards. Lesen Sie mehr darüber im neuen gugler* Nachhaltigkeitsbericht: http://www.gugler.at/unternehmen/publikationen/g-nachhaltigkeitsbericht.html