Overview
You should add some >/dev/null
, else this might show up on %post..
TARGET SOURCE FSTYPE OPTIONS / /dev/mapper/fjdata btrfs rw,relatime,attr2,inode64,noquota TARGET SOURCE FSTYPE OPTIONS /var /dev/mapper/fjdata btrfs rw,relatime,attr2,inode64,noquota TARGET SOURCE FSTYPE OPTIONS /opt /dev/mapper/fjdata btrfs rw,relatime,attr2,inode64,noquota
I do not think this will work in the context of SUSE CaaSP or openSUSE Kubic
During a transactional update (the only supported mechanism for updating or installing this package), /var is not mounted
Therefore "! findmount /var" will fail
Therefore it will try and make a /var/lib/cni subvolume with the subvolume of the transactional update
This will then be override by the /var mount on CaaSP and Kubic, but that will not have the directory created
That means in CaaSP or Kubic the entire if findmnt -t btrfs
block seems like a no-op that will not accomplish anything you need accomplished
Have you tested this on CaaSP or Kubic? can you please present the logs? Because there is no way I can see this working and would recommend we do NOT merge this.
--
Also, lines 14-15 don't make sense. You are requiring snapper if is_susecaasp, but using snappers mksubvolume on lines 26-27 regardless of whether or not is_susecaasp is true or not.
This does not address any of the issues raised in SR# 581729
(Copying here for the sake of redundancy)
I do not think this will work in the context of SUSE CaaSP or openSUSE Kubic
During a transactional update (the only supported mechanism for updating or installing this package), /var is not mounted
Therefore "! findmount /var" will fail
Therefore it will try and make a /var/lib/cni subvolume with the subvolume of the transactional update
This will then be override by the /var mount on CaaSP and Kubic, but that will not have the directory created
That means in CaaSP or Kubic the entire if findmnt -t btrfs block seems like a no-op that will not accomplish anything you need accomplished
Have you tested this on CaaSP or Kubic? can you please present the logs? Because there is no way I can see this working and would recommend we do NOT merge this.
--
Also, lines 14-15 don't make sense. You are requiring snapper if is_susecaasp, but using snappers mksubvolume on lines 26-27 regardless of whether or not is_susecaasp is true or not.
For what it's worth - This is the code Thorsten and yourself provided us to do just this! Do you have an alternative suggestion?
we were wrong :) For what it's worth, I certainly didn't think about the implications of running this as part of a transactional update where all of the findmnt's me and thorsten suggested will not work.
But the implications are there - /var will not be mounted during a transactional update
findmnt -s /var
(to read the /etc/fstab) might be a workaround
I'd really like to see some tests to make me confident this all works
Also, as this package is creating /var/lib/cni, it really needs to be defined in %files in the specfile - right now this package will be making the folder orphaned, which is really not good.
Request History
containersteam created request
autosubmit from concourse
factory-auto added opensuse-review-team as a reviewer
Please review sources
factory-auto added repo-checker as a reviewer
Please review build success
factory-auto accepted review
Check script succeeded
licensedigger accepted review
ok
namtrac accepted review
staging-bot added as a reviewer
Being evaluated by staging project "openSUSE:Factory:Staging:adi:79"
staging-bot accepted review
Picked openSUSE:Factory:Staging:adi:79
repo-checker accepted review
cycle and install check passed
staging-bot accepted review
ready to accept
staging-bot approved review
ready to accept
dimstar_suse accepted request
Accept to openSUSE:Factory