On Oct 10, 2011, at 2:22 PM, Seth Goldberg wrote:

> 
> 
> Quoting David Sechrest, who wrote the following on Mon, 10 Oct 2011:
> 
>> 
>> On Oct 10, 2011, at 11:08 AM, Glenn Lagasse wrote:
>> 
>>> * David Sechrest ([email protected]) wrote:
>>>> We are seeing the following error on a Solaris 11 b175 system while trying 
>>>> to install some
>>>> packages and need some help in resolving this.
>>>> 
>>>> I did set BE_PRINT_ERR=true in order to display the extra debug info below:
>>>> 
>>>> % pkg install pkg://solaris/runtime/java 
>>>> pkg://solaris/runtime/java/runtime64 
>>>> pkg://solaris/library/java/host-config 
>>>> pkg://solaris/system/locale/en-extra pkg://solaris/system/locale/en_us 
>>>> pkg://solaris/system/locale pkg://solaris/library/java/javadb 
>>>> pkg://solaris/system/management/service-tag 
>>>> consolidation/cacao/cacao-incorporation library/cacao 
>>>> library/cacao/cacao-crews library/cacao/cacao-glassfish 
>>>> library/cacao/cacao-me
>>>>          Packages to install:  11
>>>>      Create boot environment:  No
>>>> Create backup boot environment: Yes
>>>>           Services to change:   1
>>>> 
>>>> DOWNLOAD                                  PKGS       FILES    XFER (MB)
>>>> Completed                                11/11   2105/2105    43.9/43.9$<3>
>>>> 
>>>> be_mount_callback: failed to mount dataset rpool/ROOT/solaris-1/var at 
>>>> /tmp/.be.sfaGvh/var: directory is not empty
>>> 
>>> There's your problem.  You have a separate dataset for /var
>> 
>> Which seems to be the way text installer configures zfs for b175
>> 
>> 
>>> which beadm
>>> is trying to mount but can't because the /var directory in the BE
>>> dataset isn't empty and ZFS won't allow you to mount a dataset on a
>>> non-empty directory by default.  So, have a look in /tmp/.be.sfaGvh/var
>>> and see what's in there.
>> 
>> It's empty
>> 
>> % ls -la .be.sfaGvh/
>> total 32
>> drwx------   2 root     root         117 Oct 10 09:46 ./
>> drwxrwxrwt  17 root     sys         1199 Oct 10 14:07 ../
>> 
>> This directory was created by the pkg install command (or some sub process 
>> of it). I just
>> retried the command and if failed from the same error only this time it 
>> complained about
>> a new dir, /tmp/.be.k9aiaj, which is also empty and didn't exist prior to 
>> executing the pkg install
>> command.
>> 
>> thanks
>> Dave
>> 
>> 
>>> That will tell you what's written to that dir
>>> before the /var dataset could be mounted in the BE.
>>> 
> 
>  The beadm command tries to clean up after itself.  Try mounting the root 
> dataset for the solaris-1 BE directly, then, i.e.:
> 
>  zfs mount -o mountpoint=/a rpool/ROOT/solaris-1

There is no solaris-1 BE,

% beadm list
BE      Active Mountpoint Space  Policy Created          
--      ------ ---------- -----  ------ -------          
solaris NR     /          18.32G static 2011-10-07 10:17 

% zfs mount -o mountpoint=/a rpool/ROOT/solaris-1
cannot open 'rpool/ROOT/solaris-1': dataset does not exist

If you would like to examine the system I can send you a pointer offline.

thanks
Dave

> 
>  THEN look in /a/var and see if there's something in there -- that's what 
> would be preventing zfs from mounting the subordinate /var in there.
> 
> --S

---
David Sechrest                                     Phone:  (408)276-5800   
x15800
Systems Management                        Email:  [email protected]
Oracle Corporation




_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to