Rudder

Pull Request URL added: https://github.com/Normation/rudder-packages/pull/119

Jon, could you review it please?

Nicolas, I can't review this, because I don't understand what you're trying to do, and how it fits into other recent changes.

I recall that we added a backup mechanism recently to save uuid.hive on package uninstalls and upgrades. How does this change fit in with this? Could you explain and link to the related issue please?

Also, what is the cause that means this happens on 2.4.8, 2.6.4, 2.6.5, 2.7.1 and 2.7.2 only?

Put simply, please explain the context around this.

Jonathan CLARKE wrote:

Nicolas, I can't review this, because I don't understand what you're trying to do, and how it fits into other recent changes.

I recall that we added a backup mechanism recently to save uuid.hive on package uninstalls and upgrades. How does this change fit in with this? Could you explain and link to the related issue please?

Also, what is the cause that means this happens on 2.4.8, 2.6.4, 2.6.5, 2.7.1 and 2.7.2 only?

Put simply, please explain the context around this.

During upgrade, the previous version of pre/postuninstall is used which leads rudder-agent from the released versions 2.4.8, 2.6.4, 2.6.5, 2.7.1 and 2.7.2 to remove their /opt/rudder/etc/uuid.hive file because of #3634 et #3896.

#3915 will prevent the packages before 2.6.4, 2.6.5, 2.7.1 and 2.7.2 to remove /opt/rudder/etc/uuid.hive as the rudder-agent of these version does not contain #3634 et #3896.

This fix will replace "no uuid during upgrade" by the latest backuped one or by a new one.

This is more complex than expected. With Jon, we had a discussion because of the ordering of the pre/post script during upgrade. Indeed, as said in #3915 the order is:

[...]
   1. Run the %pre section of the RPM being installed.
   2. Install the files that the RPM provides.
   3. Run the %post section of the RPM.
   4. Run the %preun of the old package.
   5. Delete any old files not overwritten by the newer version. (This step deletes files that the new package does not require.)
   6. Run the %postun hook of the old package.
[...]

• Move the check of UUID to a shared file.
• Move the logic of the cron to a shared file.
• Install the shared file by rudder-agent package.
• Postinst of a package installed after rudder-agent (rudder-server-root seems to be a good candidate) should launch the shared file (in order to fix the issue of this ticket)
• Change logic of the cron to call the shared file: Update the system Techniques in order to modify /etc/cron.d/rudder-agent.

Pull Request updated. Jon, could you review it please ?
If validated, it will request another ticket to modify the cron entry.

Applied in changeset commit:530939f77357e20ea1da2b6414a288dfaab3baef.

Ok, this fix doesn't work on Centos ...
Cron complains that

Sep 24 18:53:37 server crond[24395]: (root) BAD FILE MODE (/etc/cron.d/rudder-agent-uuid)

PR updated with an extra fix.

Applied in changeset commit:a01181b20a0f08700f5cfaae6a483f4ef2813aeb.

Actually, this is not fixed; it cannot find the existing backup of the uuid.hive, due to a sad typo
I'm reopening the ticket

Here is what happens

 bash -x /opt/rudder/bin/check-rudder-agent 
+ set -e
+ BACKUP_DIR=/var/backups/rudder/
++ whoami
+ '[' '!' root = root ']'
+ check_and_fix_rudder_uuid
+ UUID_FILE=/opt/rudder/etc/uuid.hive
+ LATEST_BACKUPED_UUID=
+ '[' '!' -e /opt/rudder/etc/uuid.hive ']'
+ '[' -d /var/backups/rudder/ ']'
++ tail -n1
++ ls -v1 /var/backups/rudder/uuid-20131007.hive
+ LATEST_BACKUPED_UUID=/var/backups/rudder/uuid-20131007.hive
+ '[' -f /var/backups/rudder//var/backups/rudder/uuid-20131007.hive ']'
+ echo -n 'WARNING: The UUID of the node does not exist and no backup exist. A new one will be generated...'
WARNING: The UUID of the node does not exist and no backup exist. A new one will be generated...+ uuidgen
+ echo ' Done'