[Gluster-users] error in pgbench on glusterfs

[Anand Avati]

On Wed, Sep 12, 2012 at 12:05 AM, siga hiro <hirokisiga at gmail.com> wrote:

> Hi.
>
> I make the data area of postgresql in glusterfs on the server.
> And,I mounted a glusterfs area from postgres-node, and postgresql started.
> Then,It became an error when pgbench start.
> ( if fuse-mount,became error.if nfs-mount,not error.
>  mount -t glusterfs 172.22.0.181:/syncdata /syncdata -> pgbench ERROR
>  mount -t nfs -o nolock,vers=3,tcp,soft,timeo=3 172.22.0.181:/syncdata
> /syncdata -> pgbench OK
> )
>
> I found this [https://bugzilla.redhat.com/show_bug.cgi?id=810944]...
>
> Is it the bug of the kernel, or can i cope with it with glusterfs, or
> is there what or an idea would you teach?
>
>
> Node:
>   postgres node:172.22.0.180
>     OS:ScientificLinux61
>     DB:postgresql91-9.1.4-3PGDG
>
>   glusterfs node 172.22.0.181/172.22.0.182
>     OS:ScientificLinux61
>        glusterfs3.3.0-1
>
> Details:
>
> ・I make the data area of postgresql on glusterfs.
>     # gluster peer status -> 172.22.0.181
>     Number of Peers: 1
>     Hostname: 172.22.0.182
>     Uuid: 601a3cf7-8614-4b89-83e4-2dab319f3582
>     State: Peer in Cluster (Connected)
>
>     # gluster volume info
>     Volume Name: syncdata
>     Type: Replicate
>     Volume ID: c7708db1-0eb7-4a0d-a21c-24162ac3ed8f
>     Status: Started
>     Number of Bricks: 1 x 2 = 2
>     Transport-type: tcp
>     Bricks:
>     Brick1: 172.22.0.181:/home/syncdata
>     Brick2: 172.22.0.182:/home/syncdata
>
>   postgresql node
>     # mount -t glusterfs 172.22.0.181:/syncdata /syncdata
>     # ls /syncdata/pgdata/
>     PG_VERSION  global       pg_ident.conf  pg_serial    pg_tblspc
> postgresql.conf
>     base        pg_clog      pg_multixact   pg_stat_tmp  pg_twophase
> postmaster.opts
>     pg_hba.conf  pg_notify      pg_subtrans  pg_xlog      postmaster.pid
>
> ・Then, pgbench was started.
>   postgresql node
>     $ /usr/pgsql-9.1/bin/pgbench -i -s 100
>     .....
>     2012-09-12 11:53:51 ERROR:  unexpected data beyond EOF in block
> 99480 of relation base/12780/16396
>
>     $ /usr/pgsql-9.1/bin/pgbench -h 172.22.0.182  -c 100 -t 10
>     2012-09-12 11:53:51 ERROR:  unexpected data beyond EOF in block
> 99480 of relation base/12780/16396


Can you check if any of the following options fix your problem -

1) --direct-io-mode=enable (to glusterfs)
2) --attribute-timeout=0 (to glusterfs)
3) gluster volume set syncdata performance.write-behind off

Thanks,
Avati
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20120920/afe52c74/attachment.html>