Extremadura2007MeetingMinutes

We had a little meeting about some FAI related stuff at the Extremadura work meeting in october 2007. Here are the minutes from this meeting:

Issues od setup harddisk(2)

There are currently problems to "preserve the right stuff" based on disk topologies. The problems relates to the inability of having the correct and same sorting of disks and device files for them. It is there in all newer kernels - which means since woody, probably. That means, it is no problem that is newly introduced by setup harddisk2. That doesn't mean, it should not be worked on, if possible, but it means setup harddisk is no regression regarding this if it cannot handle this correctly.

Possible things to do to solve it:

• set UUIDS at first install, and keep them somehow even through new installs, set disk configs to preserve by UUID
• Store serialnumbers of the disk in the disk config file, and preserve data based on this
• look into content of disks (e.g. .diskname) to preserve based on this (e.g. with the help of faimountdisk)

Apart from that, LVM with the new tool work fine - you can:

• keep some volumes by name
• Keep all volumes or none

Stockholm has some further requirements on setup harddisk:

• It should be possible to add physical volumes to existsing groups in lvm

Henning Sprang thinks, this is advanced, and more an update topic - "setup harddisk" is something else than "change logical volume config".

A proposal for next steps to do to work a structrued way to get setup harddisks well tested, with the goal of getting it into FAI soon (maybe as default disk config tool, or in parallel with the current one):

• create a list of goals of setup harddisk 2, and check if they are met - best put at Setup_harddisks_2
• create a list of existing serious bugs that prevent putting the tools into the next stable FAI version - also at Setup_harddisks_2
• make a list of advanced wishlist things (like stockholm change stuff for lvm) - for the next version, on another wiki page - e.g. SetupHarddiskWishlist

Unit tests for FAI

Someone mentions something like "it would be good to have unit tests for FAI, but it is not possible" because the test setup is so hard.

We discuss this a little bit. The point is, Unit tests are possible. You doin't check how FAI acts on a real disk (for example with setup harddisks), but you check that the script, for a given input (which is the text output from a tool reading the disk config, and the disks geometry), the command run is the right one. The point of unit test is exactly not testoing many components together, but single ones, if the do the right thing on the right input. When looking at the whole disk after modification by setup harddisk(AND it's calls to sfdisk, or whatever), you test setup harddsk AND sfdisk, not only one.

Some comments from Henning Sprang (not from the meeting but now, when writing these minutes): There are some thoughts about the use of shunit to do such things, there is also a unit test module for perl, and there is crucible, for system level, testing. I mailed that to the list some times. There also a wiki page, where further thoughts should be collected: Testing

Testing start with the definition