[Gluster-users] gfs vs nfs: abysmal performance difference
Anand Avati
avati at gluster.com
Tue May 26 13:43:12 UTC 2009
Paolo,
Do you have a core dump on your system (in /)? Can you get us a backtrace from it?
Thanks,
Avati
----- Original Message -----
From: "Paolo Pisati" <paolo.pisati at tomato.it>
To: "Paolo Pisati" <paolo.pisati at contactlab.com>
Cc: "Anand Avati" <avati at gluster.com>, "Raghavendra G" <raghavendra at gluster.com>, gluster-users at gluster.org
Sent: Tuesday, May 26, 2009 6:11:53 PM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi
Subject: Re: [Gluster-users] gfs vs nfs: abysmal performance difference
Paolo Pisati wrote:
> Anand Avati wrote:
>
>> Can you post the client and server logs of this session?
>>
> i'm doing a parallel build test now (make -j16), later i'll try bonnie
> again.
>
> where does gluster put its logs?
>
never mind, i found the user guide:
client:
> sudo bonnie -d /mnt/gfs -m GFS
File '/mnt/gfs/Bonnie.94754', size: 104857600
Writing with putc()...done
Rewriting...done
Bonnie: drastic I/O error (/mnt/gfs/Bonnie.94754): Input/output error
/var/log/messages:
May 26 13:33:25 faustoweb sudo: piso : TTY=ttyp1 ; PWD=/usr/src ;
USER=root ; COMMAND=/usr/local/gfs/sbin/glusterfs -f
/usr/local/gfs/etc/glusterfs/glusterfs.vol -l /var/log/gfs.log -L
WARNING /mnt/gfs/
May 26 13:33:53 faustoweb sudo: piso : TTY=ttyp1 ; PWD=/usr/src ;
USER=root ; COMMAND=/usr/local/bin/bonnie -d /mnt/gfs -m GFS
May 26 13:37:30 faustoweb kernel: pid 94748 (glusterfsd), uid 0: exited
on signal 11 (core dumped)
/var/log/gfs.log:
================================================================================
Version : glusterfs 2.0.0 built on May 11 2009 09:41:53
TLA Revision : 7b2e459db65edd302aa12476bc73b3b7a17b1410
Starting Time: 2009-05-26 13:33:25
Command line : /usr/local/gfs/sbin/glusterfs -f
/usr/local/gfs/etc/glusterfs/glusterfs.vol -l /var/log/gfs.log -L
WARNING /mnt/gfs/
PID : 94747
System name : FreeBSD
Nodename : faustoweb.tomato.lan
Kernel Release : 7.2-RELEASE
Hardware Identifier: amd64
Given volfile:
+------------------------------------------------------------------------------+
1: volume client
2: type protocol/client
3: option transport-type tcp
4: option remote-host 10.0.0.1 # IP address of the remote brick
5: option remote-subvolume brick # name of the remote volume
6: end-volume
7:
8: ### Add readahead feature
9: volume readahead
10: type performance/read-ahead
11: option page-size 1MB # unit in bytes
12: option page-count 2 # cache per file = (page-count x
page-size)
13: subvolumes client
14: end-volume
15:
16: ### Add IO-Cache feature
17: volume iocache
18: type performance/io-cache
19: option page-size 256KB
20: option page-count 2
21: subvolumes readahead
22: end-volume
23:
24: ### Add writeback feature
25: volume writeback
26: type performance/write-behind
27: option aggregate-size 1MB
28: option window-size 2MB
29: option flush-behind off
30: subvolumes iocache
31: end-volume
+------------------------------------------------------------------------------+
2009-05-26 13:33:25 W [xlator.c:555:validate_xlator_volume_options]
writeback: option 'window-size' is deprecated, preferred is
'cache-size', continuing with correction
2009-05-26 13:33:25 W [glusterfsd.c:455:_log_if_option_is_invalid]
writeback: option 'aggregate-size' is not recognized
2009-05-26 13:33:25 W [glusterfsd.c:455:_log_if_option_is_invalid]
iocache: option 'page-count' is not recognized
2009-05-26 13:33:25 W [glusterfsd.c:455:_log_if_option_is_invalid]
readahead: option 'page-size' is not recognized
pending frames:
frame : type(1) op(CREATE)
patchset: 7b2e459db65edd302aa12476bc73b3b7a17b1410
signal received: 11
configuration details:db.h 1
dlfcn 1
libpthread 1
spinlock 1
extattr.h 1
st_atimespec.tv_nsec 1
package-string: glusterfs 2.0.0
server:
/var/log/messages:
May 26 12:32:19 gnode sudo: piso : TTY=ttyp0 ; PWD=/usr/home/piso ;
USER=root ; COMMAND=/usr/local/gfs/sbin/glusterfsd -f
/usr/local/gfs/etc/glusterfs/glusterfsd.vol -l /var/log/gfsd.log -L WARNING
/var/log/gfsd.log:
================================================================================
Version : glusterfs 2.0.0 built on May 11 2009 14:38:00
TLA Revision : 7b2e459db65edd302aa12476bc73b3b7a17b1410
Starting Time: 2009-05-26 12:32:19
Command line : /usr/local/gfs/sbin/glusterfsd -f
/usr/local/gfs/etc/glusterfs/glusterfsd.vol -l /var/log/gfsd.log -L WARNING
PID : 2552
System name : FreeBSD
Nodename : gnode.tomato.lan
Kernel Release : 7.2-RELEASE
Hardware Identifier: i386
Given volfile:
+------------------------------------------------------------------------------+
1: volume posix
2: type storage/posix # POSIX FS translator
3: option directory /usr/src # Export this directory
4: end-volume
5:
6: volume locks
7: type features/locks
8: option mandatory-locks on
9: subvolumes posix
10: end-volume
11:
12: volume brick
13: type performance/io-threads
14: option thread-count 8
15: subvolumes locks
16: end-volume
17:
18: volume server
19: type protocol/server
20: option transport-type tcp
21: subvolumes brick
22: option auth.addr.brick.allow * # Allow access to "brick" volume
23: end-volume
+------------------------------------------------------------------------------+
2009-05-26 12:32:19 W [xlator.c:287:_volume_option_value_validate]
brick: '8' in 'option thread-count 8' is out of range [16 - 256]
--
Paolo Pisati
Reparto Sistemi
---------
ContactLab
Via Natale Battaglia, 10
20127 Milano (Italy)
Tel. +39 02 28 31 181
Fax +39 02 70 03 02 69.
http://www.contactlab.com
More information about the Gluster-users
mailing list