Jack Schwartz wrote: > Hi Jean. > > If I understand, the problem you are seeing has to do with the fact > that the bootroot is currently built on a lofi-mounted filesystem, > correct? > > When we implement bootroot autosizing, this will change. We will do > all bootroot file population / manipulation in a directory, not in the > lofi-mounted file system, and only the bootroot_archive script will > deal with the lofi-mounted archive. This means you can snapshot the > bootroot contents during the different phases of bootroot setup. Super. So this issue is for only a short time. Do you want me to put the effort in to put out the error message? Or just let it be for now since the autosizing changes will follow shortly?
Jean > > Thanks, > Jack > > Jean McCormack wrote: >> We currently have 3 finalizer modules that create the bootroot. >> >> bootroot_initialize: creates and populates the bootroot >> bootroot_configure: Does configuration on the bootroot. This includes >> creating the smf repository. >> bootroot_archive: unmount and lofiadm -d the bootroot areas and >> compressing. >> >> After this the DC does post boot root modifications to the pkg_image >> area and creates the iso's and usb's. >> >> Unfortunately, it is impossible to rollback to the bootroot >> checkpoints after the archive step. i.e. after >> the microroot areas have been unmounted and lofiadm -d. Even if you >> go about recreating the mounts. >> Checkpointing works fine if you have not done that step. >> >> It appears we have some options on what we can do. >> >> 1) Move the ti_release_target to the end of the build. I don't even >> know if this is possible and all it does >> is narrow the window of when checkpointing the bootroot doesn't work. >> >> 2) During initialize and configure check to see if the mounts are >> there and if so rollback and everything works. >> If they are not there, display an error message telling the user they >> need to resume from the bootroot_initialize >> step. >> >> I don't see option 1 as really helping much and will proceed with >> option 2. Other solutions are welcome. >> >> Jean >> >> >> >> >> >> _______________________________________________ >> caiman-discuss mailing list >> caiman-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >> >
