[Linux-cluster] Clustering NFS4 with bests practices (bind)

GouNiNi gounini.geekarea at gmail.com
Wed Apr 25 14:06:38 UTC 2012


My question is similar with http://www.redhat.com/archives/linux-cluster/2007-April/msg00125.html but it was in 2007.
I have a cluster with HALVM + ext3 (no GFS). I need to bind some directories in my chroot NFS4 directory for various technical reaseon.

I tried many configurations but no succes. Here is one :

        <ip address="XX.XX.XX.XX/26" monitor_link="1"/>
        <lvm lv_name="lv_applis_foobar" name="lv_applis_foobar" vg_name="vg_applis_foobar"/>
        <fs device="/dev/vg_applis_foobar/lv_applis_foobar" force_unmount="1" fstype="ext3" mountpoint="/applis/foobar" name="fs_applis_foobar"/>
        <fs device="/applis/foobar" force_unmount="1" mountpoint="/exports/applis/foobar" name="bind_applis_foobar" options="bind"/>
        <nfsexport name="/exports/applis/foobar"/>
        <nfsclient fsid="100" name="exp_/exports/applis/foobar" options="rw,sync" path="/exports/applis/foobar" target="*.pma-dstage"/>
<service autostart="1" domain="data" name="files.foobar.com" nfslock="1" recovery="relocate">
        <lvm ref="lv_applis_foobar">
                <fs ref="fs_applis_foobar">
                        <fs ref="bind_applis_foobar">
                                <nfsexport ref="/exports/applis/foobar">
                                        <nfsclient ref="exp_/exports/applis/foobar"/>
        <ip ref="XX.XX.XX.XX/26"/>

Logs say :

Apr 24 17:53:08 hostname rgmanager[13264]: [fs] start_filesystem: Could not match /applis/foobar with a real device
Apr 24 17:53:15 hostanme rgmanager[11094]: start on fs "bind_applis_foobar" returned 2 (invalid argument(s))

Do you already use bind option in cluster.conf?


Jean-Daniel Bonnetot

More information about the Linux-cluster mailing list