| authors | state |
|---|---|
Josh Wilsdon <jwilsdon@joyent.com> |
predraft |
With OS-3429, SmartOS added support for the creation of the /lastexited file in the zoneroot of a zone when the zone's init process exits. This is handled by zoneadmd.
This functionality is used by vmadm for the 'exit_status' and 'exit_timestamp'
fields which are propagated to VMAPI as part of the VM objects and are then used
by sdc-docker to generate the STATUS value to display in the docker ps output.
Unfortunately there are cases where a zone is halted and this file is not written out. There will always be the possibility of CN panic, but this can also occur when a CN is rebooted.
- should we use some other method to determine exit_timestamp when we see a lastbooted file so we know the zone was booted at least once? If so: what?
- should we use a specific value for exit_status when we know a zone stopped, but we don't know what the exit status actually was?
- should vmadm provide the alternate exit_timestamp and exit_status values or should those be added at a different point in the stack?