|
Replies:
2
-
Last Post:
Jan 19, 2009 4:22 AM
by: frueauf
|
Threads:
[
Previous
|
Next
]
|
|
Posts:
241
From:
Registered:
5/29/07
|
|
|
|
[ha-clusters-discuss] Colorado: please review changes to support
iplg brand type zones with the HA Container agent
Posted:
Jan 15, 2009 5:48 AM
|
|
Hi,
Please review my code changes at
http://cr.opensolaris.org/~frueauf/Colorado-ha-ipkg-zone/
The change will enable to use zones of brand type ipkg on the
OpenSolaris binary distribution to failover on a Colorado based cluster.
If you want to read more about the ipkg brand type, I can recommend reading:
http://www.opensolaris.org/jive/thread.jspa?messageID=273528&
http://blogs.sun.com/jerrysblog/entry/zones_on_opensolaris_2008_11
In order to make the failover ipkg zone work, it is required to set the
UUID used for the global zone active boot environment equal on all nodes.
That UUID is then also set to a new property for the dataset which the
zone uses for its root path (and yes, this means moving forward, the
zone root path must be configured on a zfs dataset for a ipkg brand - no
UFS (or other file system) possible).
Only if the property value for org.opensolaris.libbe:uuid within the
active BE zfs file system matches the property value for
org.opensolaris.libbe:parentbe within the zone dataset (zbe) as being
the same UUID, the zone is allowed to boot.
This will be a documented step necessary as part of the initial
configuration.
This step needs to get repeated, if a new active boot environment gets
created, specifically also after an upgrade.
I did decide against coding a validate step to check for this, since
that validation is done when the zone tries to boot. The boot fails if
the zone validation fails.
If you are interested in the rather complex code, have a look at
/usr/lib/brand/ipkg/common.ksh - duplicating the code which uses
internal interfaces within the agent did not look useful to me.
Note that I can also not simply automatically adjust the UUID, since the
UUID is used to distinguish if a zone got upgraded and therefor can run
on a specific boot environment. The agent can not have knowledge about
the history of UUIDs used accross nodes when upgrades happen.
That is the reasoning for deciding to have a documented manual step at
initial configuration time and after upgrades.
This review expires at Tuesday, 20. January 2009, COB MET.
Greets
Thorsten
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Sitz der Gesellschaft:
Sun Microsystems GmbH, Sonnenallee 1, D-85551 Kirchheim-Heimstetten
Amtsgericht Muenchen: HRB 161028
Geschaeftsfuehrer: Thomas Schroeder, Wolfgang Engels, Dr. Roland Boemer
Vorsitzender des Aufsichtsrates: Martin Haering
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_______________________________________________
ha-clusters-discuss mailing list
ha-clusters-discuss at opensolaris dot org
http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss
|
|
|
Posts:
58
From:
US
Registered:
4/17/07
|
|
|
|
Re: [ha-clusters-discuss] Colorado: please review changes to
support iplg brand type zones with the HA Container agent
Posted:
Jan 15, 2009 3:50 PM
in response to: frueauf
|
|
Thorsten,
I have one comment on the user-visible text:
functions.ksh
- In an SCMSGS explanation, change "...But the Zone is configured as a
different brand type (i.e. lx) according to zonecfg info" to "But the
zone is configured as a different brand type, for example lx, according
to zonecfg information".
Otherwise it looks fine.
Lisa Shepherd
Sun Cluster Technical Publications
"We're the M in RTFM"
On 01/15/09 05:48, Thorsten Frueauf wrote:
> Hi,
>
> Please review my code changes at
> http://cr.opensolaris.org/~frueauf/Colorado-ha-ipkg-zone/
>
> The change will enable to use zones of brand type ipkg on the
> OpenSolaris binary distribution to failover on a Colorado based cluster.
>
> If you want to read more about the ipkg brand type, I can recommend reading:
>
> http://www.opensolaris.org/jive/thread.jspa?messageID=273528&
> http://blogs.sun.com/jerrysblog/entry/zones_on_opensolaris_2008_11
>
> In order to make the failover ipkg zone work, it is required to set the
> UUID used for the global zone active boot environment equal on all nodes.
>
> That UUID is then also set to a new property for the dataset which the
> zone uses for its root path (and yes, this means moving forward, the
> zone root path must be configured on a zfs dataset for a ipkg brand - no
> UFS (or other file system) possible).
>
> Only if the property value for org.opensolaris.libbe:uuid within the
> active BE zfs file system matches the property value for
> org.opensolaris.libbe:parentbe within the zone dataset (zbe) as being
> the same UUID, the zone is allowed to boot.
>
> This will be a documented step necessary as part of the initial
> configuration.
>
> This step needs to get repeated, if a new active boot environment gets
> created, specifically also after an upgrade.
>
> I did decide against coding a validate step to check for this, since
> that validation is done when the zone tries to boot. The boot fails if
> the zone validation fails.
>
> If you are interested in the rather complex code, have a look at
> /usr/lib/brand/ipkg/common.ksh - duplicating the code which uses
> internal interfaces within the agent did not look useful to me.
>
> Note that I can also not simply automatically adjust the UUID, since the
> UUID is used to distinguish if a zone got upgraded and therefor can run
> on a specific boot environment. The agent can not have knowledge about
> the history of UUIDs used accross nodes when upgrades happen.
>
> That is the reasoning for deciding to have a documented manual step at
> initial configuration time and after upgrades.
>
> This review expires at Tuesday, 20. January 2009, COB MET.
>
> Greets
> Thorsten
>
_______________________________________________
ha-clusters-discuss mailing list
ha-clusters-discuss at opensolaris dot org
http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss
|
|
|
|
|
Posts:
241
From:
Registered:
5/29/07
|
|
|
|
Re: [ha-clusters-discuss] Colorado: please review changes to
support iplg brand type zones with the HA Container agent
Posted:
Jan 19, 2009 4:22 AM
in response to: lisashep
|
|
Hi Lisa et al,
Thanks for your review. I have incorporated your suggestion and changed
it on other occasions as well.
The updated webrev is at
http://cr.opensolaris.org/~frueauf/Colorado-ha-ipkg-zone-2/
Greets
Thorsten
Lisa dot Shepherd at Sun dot COM wrote:
> Thorsten,
>
> I have one comment on the user-visible text:
>
> functions.ksh
> - In an SCMSGS explanation, change "...But the Zone is configured as a
> different brand type (i.e. lx) according to zonecfg info" to "But the
> zone is configured as a different brand type, for example lx, according
> to zonecfg information".
>
> Otherwise it looks fine.
>
> Lisa Shepherd
> Sun Cluster Technical Publications
> "We're the M in RTFM"
>
>
>
> On 01/15/09 05:48, Thorsten Frueauf wrote:
>> Hi,
>>
>> Please review my code changes at
>> http://cr.opensolaris.org/~frueauf/Colorado-ha-ipkg-zone/
>>
>> The change will enable to use zones of brand type ipkg on the
>> OpenSolaris binary distribution to failover on a Colorado based cluster.
>>
>> If you want to read more about the ipkg brand type, I can recommend
>> reading:
>>
>> http://www.opensolaris.org/jive/thread.jspa?messageID=273528&
>> http://blogs.sun.com/jerrysblog/entry/zones_on_opensolaris_2008_11
>>
>> In order to make the failover ipkg zone work, it is required to set
>> the UUID used for the global zone active boot environment equal on all
>> nodes.
>>
>> That UUID is then also set to a new property for the dataset which the
>> zone uses for its root path (and yes, this means moving forward, the
>> zone root path must be configured on a zfs dataset for a ipkg brand -
>> no UFS (or other file system) possible).
>>
>> Only if the property value for org.opensolaris.libbe:uuid within the
>> active BE zfs file system matches the property value for
>> org.opensolaris.libbe:parentbe within the zone dataset (zbe) as being
>> the same UUID, the zone is allowed to boot.
>>
>> This will be a documented step necessary as part of the initial
>> configuration.
>>
>> This step needs to get repeated, if a new active boot environment gets
>> created, specifically also after an upgrade.
>>
>> I did decide against coding a validate step to check for this, since
>> that validation is done when the zone tries to boot. The boot fails if
>> the zone validation fails.
>>
>> If you are interested in the rather complex code, have a look at
>> /usr/lib/brand/ipkg/common.ksh - duplicating the code which uses
>> internal interfaces within the agent did not look useful to me.
>>
>> Note that I can also not simply automatically adjust the UUID, since
>> the UUID is used to distinguish if a zone got upgraded and therefor
>> can run on a specific boot environment. The agent can not have
>> knowledge about the history of UUIDs used accross nodes when upgrades
>> happen.
>>
>> That is the reasoning for deciding to have a documented manual step at
>> initial configuration time and after upgrades.
>>
>> This review expires at Tuesday, 20. January 2009, COB MET.
>>
>> Greets
>> Thorsten
>>
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Sitz der Gesellschaft:
Sun Microsystems GmbH, Sonnenallee 1, D-85551 Kirchheim-Heimstetten
Amtsgericht Muenchen: HRB 161028
Geschaeftsfuehrer: Thomas Schroeder, Wolfgang Engels, Dr. Roland Boemer
Vorsitzender des Aufsichtsrates: Martin Haering
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_______________________________________________
ha-clusters-discuss mailing list
ha-clusters-discuss at opensolaris dot org
http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss
|
|
|
|
|
