Hi John,

I like this idea - it would be clear solution for the problem.
Is it possible to manage custom parameters with standard CLI
commands (these are used by the installer for manipulating ZFS
entities) - zpool(1M), zfs(1M), ... ?

I am asking, since I haven't found information about setting
custom parameters neither in man pages nor in
"Solaris ZFS Administration Guide" available on opensolaris.org,
I have probably missed it.

Thank you,
Jan


John Langley wrote:
> What about setting a custom parameter on rpool when you create it and 
> then changing the value after a successful install.  Then you can test 
> for the value of the parameter and know if it was a failed install.
>
> John
>
> On Mon, Aug 18, 2008 at 11:04 AM, Jan Damborsky <[EMAIL PROTECTED] 
> <mailto:[EMAIL PROTECTED]>> wrote:
>
>     Hi ZFS team,
>
>     I am currently working on fixes for couple of bugs in OpenSolaris
>     Caiman
>     installer and since they are related to the ZFS, I would like to
>     kindly
>     ask you, if you could please help me to understand, if the issues
>     encountered and mentioned below are known (somebody works on them or
>     if they are already fixed), or if some workarounds might be used.
>
>     Also, please let us know if there is possibility that other approach
>     (like other/new API, command, subcommand) might be used in order to
>     solve the problem.
>
>     Any help/suggestions/comments are much appreciated.
>     Thank you very much,
>     Jan
>
>
>     Let me describe the main problem we are trying to address by the fix
>     which is tracked by following bug:
>
>     1771 Install will fail if rpool ZFS pool already exists
>
>     When OpenSolaris starts process of installation, it creates ZFS
>     root pool
>     "rpool" on the target, which is then populated with appropriate
>     ZFS datasets
>     and ZFS volumes (used for swap and dump) - then Solaris is
>     installed there.
>
>     If installer exits either due to some kind of failure or user
>     intervention,
>     we would like to make sure, it can be restarted again. Because of
>     this,
>     we need to destroy all allocated resources, which in ZFS case means we
>     need to release "rpool", so that installer could use it again.
>
>     The problem is that we don't know, if "rpool" was just created by the
>     installer or imported by user (for example other Solaris instance
>     might
>     exist
>     on other disk). If latter is the case, we don't want to destroy
>     the pool.
>
>     We were trying to address this problem in following ways, but none
>     of them
>     worked for us due to the issues encountered:
>
>     [1] Installing into "temporary" pool
>     ====================================
>     Installer created "rpool_tmp" for installation and when installation
>     successfully finished, it was "renamed" to "rpool".
>     If installer failed during installation, we knew that we could safely
>     remove "rpool_tmp", since this was only temporary pool and thus
>     couldn't
>     contain valid Solaris instance.
>
>     We were "renaming" the pool in following way:
>
>     # zpool create -f rpool_tmp <device>
>     [...installation...]
>     # pool_id=`zpool list -H -o guid rpol_tmp`
>     # zpool export -f rpool_tmp
>     # zpool import -f $pool_id rpool
>
>     However, we encountered problem that sometimes root "/rpool" dataset
>     was not mounted due to some reason and we failed to access it. Please
>     see more details in bug report
>
>     1350 Install without menu.lst
>
>     [2] Exporting "rpool" if present on the system
>     ==============================================
>     When installer started, it checked if "rpool" is present and
>     in that case it exported it.
>
>     # zpool export -f rpool
>
>     The problem is that this operation made "rpool" unbootable:
>
>     1365 zfs import/export will modify the boot up information of ZFS
>
>     So if user had some Solaris instance on that pool, it was not able
>     to boot it anymore (without importing the pool).
>
>     _______________________________________________
>     zfs-discuss mailing list
>     zfs-discuss@opensolaris.org <mailto:zfs-discuss@opensolaris.org>
>     http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>
>

_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Reply via email to