> Michael Tautschnig <m...@debian.org> wrote: > > Hi, > > > If done from within setup-storage, would it be as simple as to write all the > > parted commands to a separate file (as well)? If so, wouldn't it be best to > > Nope, not that simple :) > > I want one script per physical device, and the script must include any > command that might be needed to recreate the partition table, > filesystems and (I don't need that at the moment) LVM volumes. mdadm > commands aren't needed in this context. > > But that script must come out identical regardless of preserved > partitions during installation (that requires some effort). >
But things shouldn't be too bad, at least given the current code (Toomas Tamm doesn't like that :-) ) - even preserved partitions will be (re-)created by parted commands. Ok, filesystems require some extra work, but then again I'm not sure whether you actually want/need that - at least not in context of failing RAID devices. For LVM the pvcreate commands will have to be executed. > > really just print all the commands to be executed *such that this command > > script > > ends up in the logs*? To me, logs sound like a good place for keeping this > > kind > > of backup-information. If it really is just about logging the commands, the > > The goal is to have scripts in /tmp/fai that I can copy to /root or > somewhere else that is safe (RAID1) *on the machine* so the sysadmins > can just pull the failed drive, instert the new one, run the script, > rebuild the RAID. > > I have no guarantee the install logs will be available 4 years after the > deployment when disks will fail :) > > > question is whether all commands should be logged, or just a specific > > subset. If > > the latter is the case, we could extend the setup-storage code that is used > > to > > create commands to take a flag to enable or disable logging of the specific > > command. > > It needs a bit more than that, as there are some mods needed so that > preserving a partition still emits the necessary parted command in the > script. > > So you need at least 2 flags: log command and execute command. > As said above: Should mostly just be the logging, and maybe do it as follows: - Add a function that does logging - Add a flag to push_command that causes the logging function to be called - If commands are not pushed because of preserve being in effect, call the logging function directly. Best, Michael
pgpWfLcQqDkpy.pgp
Description: PGP signature