Contributed by: Jan Pechanec

The Live Zone Reconfiguration, LZR for short, was introduced in Oracle Solaris 11.3. It allows for changing a configuration of a running zone without a need to reboot it. More precisely, it provides a way to:

  • reconfigure a running zone to match persistent configuration of a zone maintained through zonecfg(8)
  • edit the live configuration of a running zone instead of a persistent configuration on a stable storage

Up to Solaris 11.3, the LZR supported resources and properties like anet, capped-memory, pool, and others. For the complete list, see the solaris(5) and solaris-kz(5) manual pages on 11.3 as there are differences in what is supported by each brand. Please note that those manual pages were renamed in 11.4 Beta to zones_solaris(7) and zones_solaris-kz(7).

With the release of Oracle Solaris 11.4 Beta, the LZR now newly supports reconfiguration of dataset resources for the “solaris” branded zones — aka non-global zones (NGZs), or native zones.

This new feature in 11.4 will allow you to add and remove datasets to solaris branded zones on the fly, whereas such datasets will show up within the zone as ZFS pools that may be imported and exported using an established zpool(8) command.

In order to provide a coherent user interface and a good administrative experience, the decision was to use already existing NGZ command line interface to import a new virtual ZFS pool that has been made available to the running NGZ by the GZ administrator via adding a dataset resource using the LZR, and also to remove a dataset resource out of a running NGZ when no longer needed.

A short example will illustrate how to use this feature via applying a modified persistent configuration. First, create the dataset in the global zone, modify the native zone’s configuration, and apply the modifications:

global-zone# zfs create rpool/lzr1

global-zone# zonecfg -z tzone1
zonecfg:tzone1> add dataset 
zonecfg:tzone1:dataset> set name=rpool/lzr1
zonecfg:tzone1:dataset> end
zonecfg:tzone1> commit
zonecfg:tzone1> exit

global-zone# zoneadm -z tzone1 apply
zone 'tzone1': Checking: Adding dataset name=rpool/lzr1
zone 'tzone1': Applying the changes

Now log in to the tzone1 zone and import the newly added ZFS pool:

global-zone# zlogin tzone1

tzone1:~# zpool list
NAME   SIZE  ALLOC   FREE  CAP  DEDUP  HEALTH  ALTROOT
rpool  127G  32.0G  95.0G  25%  1.00x  ONLINE  -

tzone1:~# zpool import	# will show pools available for import
  pool: lzr1
    id: 11612286362086021519
 state: ONLINE
action: The pool can be imported using its name or numeric identifier.
config:

        lzr1    ONLINE
          c1d0  ONLINE

tzone1:~# zpool import lzr1

tzone1:~# zpool list
NAME   SIZE  ALLOC   FREE  CAP  DEDUP  HEALTH  ALTROOT
lzr1   127G  32.0G  95.0G  25%  1.00x  ONLINE  -
rpool  127G  32.0G  95.0G  25%  1.00x  ONLINE  -

tzone1:~# touch /lzr1/hello
tzone1:~# ls -l /lzr1/hello
-rw-r--r--   1 root     root           0 Jan 29 09:51 /lzr1/hello

The size of the imported lzr1 pool is same as how much space is available in the ZFS dataset rpool/lzr1 within the GZ.

Now trying to remove the dataset while it is still imported in the zone:

global-zone# zonecfg -z tzone1
zonecfg:tzone1> remove dataset 0
zonecfg:tzone1> commit
zonecfg:tzone1> exit

global-zone# zoneadm -z tzone1 apply
zone 'tzone1': Checking: Removing dataset name=rpool/lzr1
zone 'tzone1': error: cannot remove dataset 'rpool/lzr1': not in exported state
zoneadm: zone tzone1: failed to apply the configuration: zoneadmd(8) returned an error 1 (unspecified error)
global-zone# echo $?
1

As you would probably expect, one cannot remove a dataset resource while the ZFS pool is in use within the zone. So, just redo after exporting the pool first:

tzone1:~# zpool export lzr1
tzone1:~# echo $?
0

global-zone# zoneadm -z tzone1 apply
zone 'tzone1': Checking: Removing dataset name=rpool/lzr1
zone 'tzone1': Applying the changes

Alternatively, you could only update the in-memory configuration via “zonecfg -z tzone1 -r“. The configuration is applied on commit:

global-zone# zonecfg -z tzone1 remove dataset 0
global-zone# zonecfg -z tzone1 apply
...
global-zone# zonecfg -z tzone1 -r
zonecfg:tzone1> add dataset
zonecfg:tzone1:dataset> set name=rpool/lzr1
zonecfg:tzone1:dataset> end
zonecfg:tzone1> commit
zone 'tzone1': Checking: Adding dataset name=rpool/lzr1
zone 'tzone1': Applying the changes

Within the tzone1 zone, you would now import the pool same way as before.