I have been experimenting on the above system, using ldmp2v to create "clones" of my physical systems as LDoms on the server when there was an unscheduled power outage. After the system came back up I had lost my LDoms, although the ZFS Volume backends still exist on the system.
Now, I am ready to accept the blame; that I should have saved a configuration so that the system knew what to do to bring them back up. Except that I haven't come across anything in the documentation to say that I had to do this; or what was required to do so.
Now it is possible that I should have used
ldm list-constraints -x domain
to save the LDom configuration, but where do I save it to?
Also, can I build new LDoms manually and attach the existing backends to them, or should I just delete the backends and start again?
As an aside, is it possible to mount and read these on the server?
Looks like you only lost the configuration, not the actual ldoms and data inside.
This is strange, even when power off occurred i never lost all configuration.
Just as an aside, you can always import the zpools into the service domain and check them if you have to.
As there likely be a clash in the names, for the rpools - you'll likely have to use the numeric id to import or you can just import to a different name as in;
Funny story. While trying to figure out what happened I noticed that the output of ldm list produced the factory default of 128 VCPUs and 129280M memory, rather than the values I used when following the instructions here (Configuring the Control Domain). As I had pasted what I had done in a personal wiki I checked what I did and noticed this:
sog01(31)$ sudo ldm add-config initial
Error: Operation failed because a configuration
named "initial" already exists on the system controller.
Before being able to save a new configuration with
this name the existing one must be removed
sog01(32)$ ldm list-config
factory-default
initial [next poweron]
I missed that error at the time. So yesterday I followed the instructions again, this time changing the add-config command to
sudo ldm add-config main
and then checked that the configuration stuck by shutting down and literally pulling the plug out of the wall. Now:
sog01(13)$ ldm list-config
factory-default
initial
main [current]
This is strange behavior.
Such systems should never lose configuration in that manner (reset to factory).
This would indicate issues with ILO, since configuration is saved there, and retrieved by the hypervisor during boot.
Did you ever issue save configuration command ldm add-spconfig <friendly_name_date> on that system ?
Check out firmware version and update if you can (since you have downtime now).
As for the hypervisor configuration, you will need to reconfigure primary with couple of cores / GB of ram, not leave it configured with all machines capacity assigned to primary (and in your case, control domain as well).
Other resources (CPU,RAM) are left out to ldoms.
Consider limiting the ZFS ARC on primary domain using set user_reserve_hint_pct=80 in /etc/system.
Should I be running that every time I add/modify an LDom? The documentation is pretty poor and appears to be aimed at those who know what they are doing.
It's a while since I posted into this thread. I just wanted to say thanks to all those who replied. I have my LDoms back, and hopefully won't fall into the same trap again.