[Linux-cluster] postgres cluster with RHEL5
Terry
td3201 at gmail.com
Fri Feb 1 19:14:37 UTC 2008
On Feb 1, 2008 10:30 AM, Terry <td3201 at gmail.com> wrote:
> On Feb 1, 2008 10:09 AM, Lon Hohberger <lhh at redhat.com> wrote:
> > On Thu, 2008-01-31 at 23:10 -0600, Terry wrote:
> >
> > > <service autostart="1" domain="failover00" exclusive="0"
> > > name="database" recovery="relocate">
> > > <postgres-8 ref="dssystem2">
> > > <fs ref="db00"/>
> > > <ip ref="192.168.100.207"/>
> > > </postgres-8>
> > > </service>
> >
> > That's not going to work; the dependencies are backwards (assuming
> > postgresql is trying to use the fs and ip). Try:
> >
> > <service autostart="1" domain="failover00" exclusive="0"
> > name="database" recovery="relocate">
> > <fs ref="db00"/>
> > <ip ref="192.168.100.207"/>
> > <postgres-8 ref="dssystem2"/>
> > </service>
> >
> > -- Lon
>
> I spoke incorrectly. The shared storage is an iSCSI volume with ext3.
> NFS is another cluster service. Sorry.
>
Thanks for all of your replies. I think I have a catch 22 situation.
Here is my (corrected) config:
<resources>
<ip address="192.168.100.206" monitor_link="1"/>
<fs device="/dev/sdc1" force_fsck="0"
force_unmount="1" fsid="42585" fstype="ext3" mountpoint="/db00"
name="db00" self_fence="1"/>
<postgres-8
config_file="/var/lib/pgsql/data/postgresql.conf" name="dssystem"
postmaster_user="postgres" shutdown_wait="0"/>
</resources>
<service autostart="1" domain="failover00"
exclusive="0" name="database" recovery="relocate">
<fs ref="db00"/>
<ip ref="192.168.100.207"/>
<postgres-8 ref="dssystem"/>
</service>
Here is a listing of /db00:
[root at nfs00a ~]# ls /db00/
backups data pgstartup.log
[root at nfs00a ~]# ls /db00/data/
base lost+found pg_hba.conf pg_log pg_subtrans
pg_twophase pg_xlog postmaster.opts
global pg_clog pg_ident.conf pg_multixact pg_tblspc
PG_VERSION postgresql.conf
You can see that my postgres config lives on this iscsi volume which
needs to be mounted as part of the database service. When I attempt
to start the database service, it acts like it starts, there are no
logs but it doesn't start, nor does the /db00 volume get mounted.
What am I missing? I want to stay as close to standard as possible
with regards to postgres and directory locations so I created a
symlink:
lrwxrwxrwx 1 root root 5 Feb 1 12:32 /var/lib/pgsql -> /db00
I hope this paints a better picture of what I am doing and ultimately,
where I am failing.
Thanks all!
More information about the Linux-cluster
mailing list