On Fri, 26 Feb 2021 19:39:19 +0000 Dave Howorth <dave@howorth.org.uk> wrote:
On Fri, 26 Feb 2021 14:48:04 +0100 Bengt Gördén <bengan@bag.org> wrote:
On 2021-02-26 12:24, Carlos E. R. wrote:
The problem is that this procedure doesn't consider packages that can be installed from more than one repository.
I use this one
zypper --xmlout search --installed-only -v > /tmp/somefile.xml
A few years back I wrote this down but never used it. It's supposed to be able to do quite a lot of things in regards to migrating systems. http://machinery-project.org/ https://software.opensuse.org/package/machinery
A warning to anybody trying to use this!
I just installed it and ran inspect.
It appears it's filled my disk and crippled my system. I haven't finished investigating yet. More later.
OK. I've finished my investigation for now. I ran host:~ # machinery inspect -x localhost and went for dinner. Big mistake. It ran and created a directory ~root/.machinery In that directory it created a log and lots of other files that are presumably the information it is recording. When those files got to 6.8GB they filled my disk and it seems that machinery just abandoned its run at that point without logging anything to file or to the screen. The good news is that I came back after dinner to find a system that was almost but not quite unusable. systemd-journald was logging lots of messages about the inability to do things because the disk was full, but fortunately the system was still responsive. So I discovered the .machinery directory, copied it to another machine and then deleted it and now things seem to be back to normal. So far I'm less than impressed by the machinery package and the packaging. The installation was a pain, since it wants to use single-click installation, which we all agree (I think) is a bad idea. Then the manual procedure it suggests is to install the main repo and update repo that you already have. After I cleaned up its efforts, I just installed it from yast. Doh! Why didn't it suggest that in the first place! Then of course the run. Fine if it needs a load of space, but it would be sensible to let users know beforehand, and ideally provide an estimate when it starts. And when it does run out of disk before it finishes, wouldn't it make sense to log that problem with its dying breath rather than trying to store whatever was the last thing it wrote? In fact, if it runs out of disk before finishing then why doesn't it just delete everything it wrote, since I guess what I've carefully saved is entirely useless. And exactly how much disk does it need? PS I've no idea what it's actually trying to do. It has saved lots of 'changed_managed_files' that are kernel drivers. I generally don't play with those! Most of the space it has used seems to be taken up with unmanaged_files in home. Quite why it thinks it needs those to manage system upgrades etc I'm not sure. Not yet convinced!