[Ocfs2-users] OCFS2 Installation woes

Mon Oct 17 17:40:30 CDT 2005

I seem to have some system stability problems...

After installing RHEL4 Update-2 (on Dell 2850 server, dual CPU, RAID, 
4GB) and OCFS2 1.0.7 and getting a volume formatted and mounted I left 
for the weekend. The server was locked up Monday AM when I came in - no 
response to ping, "no video sync" reported by the KVM switch.

I attempted to reboot and sometimes would boot, but most often would go 
through all the motions, up to the HAL loading, then instead of the 
X-Windows login screen would get some random colors at the top of the 
screen.
Nothing seems to get reported to the /var/log/messages file.

I tried disabling the OCFS2 service and mount, but still have stability 
problems at boot with X.
I disabled X (set inittab to use run level 3) and I can boot with or 
without OCFS2.
But, without OCFS2 I can run the "startx" and have the X-Windows 
environment come up.
If OCFS2 is running (loaded + mounted) and I run startx I get the crash 
with the colors at the top of the screen.

Suggestions?

thanks,
--Peter

Sunil Mushran wrote:
> /etc/ocfs2/cluster.conf holds list of nodes in the cluster.
> /etc/sysconfig/o2cb holds the "default" cluster name which
> is picked by the o2cb service.
>
> The latter is updated during:
> /etc/init.d/o2cb configure
>
> Appears the cluster name in the two do not match.
>
> What difference did you notice?
>
> Peter Sylvester wrote:
>
>> Here is some feedback on my OCFS2 installation...
>>
>> You were correct about selinux.  I was able to disable it by 
>> specifying "SELINUX=disabled" in /etc/selinux/config and rebooting. I 
>> was then able to run the console tool correctly.
>>
>> I did stumble on another issue, though.  It appears that the console 
>> tool hard codes the cluster name as "ocfs2", but the cluster service 
>> seems to be looking for cluster "racdb", and errors out when 
>> attempting to bring online.  I hand edited the cluster.conf file and 
>> changed the cluster name to "racdb" and everything then appeared to 
>> work.
>>
>> Also, I noticed that the cluster.conf configuration is a bit 
>> different from that shown in the User's Guide documentation.
>>
>> I was able to mkfs.ocfs2 and mount the partition and it appears to 
>> work.  I'll plan on doing some testing next week to stress it out a bit.
>>
>> --Peter
>>
>> Sunil Mushran wrote:
>>
>>> You appear to have selinux enabled which is missing policies for 
>>> configfs/ocfs2.
>>> We are still investigating the issue, but one quick solution would 
>>> be to disable selinux. :)
>>>
>>> Peter Sylvester wrote:
>>>
>>>> I've got a fresh RHEL AS 4-U2 installation on a Dell PE2850 server.
>>>>
>>>> I downloaded and installed the latest RPMs:
>>>> ocfs2-2.6.9-22.ELsmp-1.0.7-1.i686.rpm
>>>> ocfs2-tools-1.0.2-1.i386.rpm
>>>> ocfs2console-1.0.2-1.i386.rpm
>>>>
>>>> I was able to start the console, but when I try to run 
>>>> cluster->configure_nodes, I get the following error message:
>>>> Could not start cluster stack. This must be resolved before any 
>>>> OCFS2 filesystem can be mounted.
>>>>
>>>> I tried to start the service via the command line:
>>>> /etc/init.d/o2cb load
>>>>
>>>> And got the error:
>>>> Mounting configfs filesystem at /config: mount: block device 
>>>> configfs is write-protected, mounting read-only
>>>> mount: cannot mount block device configfs read-only
>>>> Unable to mount configfs filesystem
>>>> Failed
>>>>
>>>> I tried mounting /config myself via the following, which worked 
>>>> (not sure if thats how it *should* be done though):
>>>> mount -t configfs none /config
>>>>
>>>> Now o2cb load puts out the following message:
>>>> Loading module "ocfs2_nodemanager": Unable to load module 
>>>> "ocfs2_nodemanager"
>>>> Failed
>>>>
>>>> And an o2cb status shows:
>>>> Module "configfs": Loaded
>>>> Filesystem "configfs": Mounted
>>>> Module "ocfs2_nodemanager": Not loaded
>>>> Module "ocfs2_dlm": Not loaded
>>>> Module "ocfs2_dlmfs": Not loaded
>>>> Filesystem "ocfs2_dlmfs": Not mounted
>>>>
>>>> I tried creating a /etc/ocfs2/cluster.conf myself (using vi), but 
>>>> it did not seem to have any effect.
>>>> I tried a fresh reboot but that did not seem to change anything.
>>>> Also note that all commands were run as root, and I am just trying 
>>>> to get a single node with local disk working.
>>>>
>>>> There were some error messages reported in /var/log/messages, of 
>>>> the form:
>>>> Oct 13 17:02:56 dblinux1 kernel: SELinux: initialized (dev 
>>>> configfs, type configfs), not configured for labeling
>>>> Oct 13 17:02:56 dblinux1 kernel: audit(1129237376.191:5): avc:  
>>>> denied  { mount } for  pid=14922 comm="mount" name="/" dev=configfs 
>>>> ino=70286 scontext=root:system_r:initrc_t 
>>>> tcontext=system_u:object_r:unlabeled_t tclass=filesystem
>>>>
>>>> Also have some errors of form:
>>>> Oct 13 18:03:49 dblinux1 dbus: Can't send to audit system: USER_AVC 
>>>> pid=2587 uid=81 loginuid=-1 message=avc:  denied  { send_msg } for  
>>>> scontext=user_u:system_r:unconfined_t 
>>>> tcontext=user_u:system_r:initrc_t tclass=dbus
>>>>
>>>> And this one:
>>>> Oct 13 17:46:36 dblinux1 kernel: OCFS2 Node Manager 1.0.7 Wed Oct 
>>>> 12 13:18:42 PDT 2005 (build 6cb35edfedddf6b4d606b95f2579cb39)
>>>> Oct 13 17:46:36 dblinux1 kernel: audit(1129239996.953:8): avc:  
>>>> denied  { mount } for  pid=3903 comm="modprobe" name="/" 
>>>> dev=configfs ino=10949 scontext=root:system_r:initrc_t 
>>>> tcontext=system_u:object_r:unlabeled_t tclass=filesystem
>>>> Oct 13 17:46:36 dblinux1 kernel: nodemanager: Registration returned 
>>>> -13
>>>> Oct 13 17:46:36 dblinux1 modprobe: FATAL: Error inserting 
>>>> ocfs2_nodemanager 
>>>> (/lib/modules/2.6.9-22.ELsmp/kernel/fs/ocfs2/ocfs2_nodemanager.ko): 
>>>> Permission denied
>>>>
>>>> Any clues as to what I should do I try from here????
>>>>
>>>> thanks,
>>>> --Peter
>>>>
>>>> _______________________________________________
>>>> Ocfs2-users mailing list
>>>> Ocfs2-users at oss.oracle.com
>>>> http://oss.oracle.com/mailman/listinfo/ocfs2-users
>>>
>>>
>>
>> _______________________________________________
>> Ocfs2-users mailing list
>> Ocfs2-users at oss.oracle.com
>> http://oss.oracle.com/mailman/listinfo/ocfs2-users
>