Disaster recovery attempts keep failing

[dnutan (Marc)]

Try the yum command without the .rpm packages extension

[fasttech]

@dnutan

Try the yum command without the .rpm packages extension

dammit… thank you sir.

[fasttech]

@giacomo my first attempt failed.

I… installed the two rpm’s on a current vm with samba dc, backup, nethserver some users/groups and files and *** php scl set to global 5.6 *** , rebooted, executed a backup.

Then started a fresh vm, updated to rc3… now the installation of the two rpm’s also installed their dependencies… this installed samba dc and backup config even before I installed backup… installed backup data, did not configure samba, configured backup, rebooted… fail after restore attempt, lost shell access, only console access.

[fasttech]

@giacomo I suppose, I should not have installed the new nethserver dc rpm in the fresh install… or… hmmmmm… I cornfussssing myself now, how can I test this if restore is going to pull from the standard repo… hmmmmm…

[giacomo (Giacomo Sanchietti)]

The only way to test it on a clean machine is enabling the testing repo: you need to set enabled=1 inside the NethServer.repo file.

[davidep (Davide Principi)]

@fasttech

now that we’re on github, we don’t “take ownership” of the issue if we’re testing? I don’t see any way to do so.

Your user account must be member of IssuesTeam on github. Now, you’ve been invited!

[fasttech]

@giacomo

The only way to test it on a clean machine is enabling the testing repo: you need to set enabled=1 inside the NethServer.repo file.

So, I don’t know where that is.
I’m guessing from my searching this might be it; /etc/yum.repos.d/NethServer.repo idk.
I don’t know if I should set base and updates to 0.

[fasttech]

@giacomo

I also wonder if I screwed up the backup by applying those rpms to the vm and executing a backup.

[giacomo (Giacomo Sanchietti)]

@fasttech

I’m guessing from my searching this might be it; /etc/yum.repos.d/NethServer.repo idk.

Yes it is.
You should simply set enabled=1 under [nethserver-testing] section.

Take your time, there is no rush: in the meanwhile we are testing the fix internally

[fasttech]

@davidep @giacomo I failed again attempting to restore the dc vm. I’ll leave the machine running if you have any log searches you’d like me to run from the console, that’s all that’s available… sshd is dead. You can see from the yum log that both the packages in testing are installed.

[davidep (Davide Principi)]

Thanks form reporting it. I hit the same problem today, and this confirm my thoughts. I’ll work on this tomorrow!

@fasttech

sshd is dead

It could be a firewall cfg issue. Run

 shorewall stop

[fasttech]

@davidep

It could be a firewall cfg issue. Run

shorewall stop

Yes, you’re correct, sshd was not dead, stopping shorewall gave me access with putty from another machine.

[fasttech]

@davidep
fyi, I just got an email from this server that a backup completed… so I logged into the web gui… successfully, though the domain, users, etc, are broken, some modules were restored like nextcloud, others not like the php scl… maybe that repo wasn’t restored, perhaps killing shorewall gave access to the admin-httpd. That also means I have to remember to get a good backup before I try another restore test.

[davidep (Davide Principi)]

A new set of RPMs is ready for testing! /cc @quality_team

Two issues should now be fixed

• Avoid applying a bad firewall configuration to prevent the admin from locking himself out
• Backup/Restore the Samba secrets.tdb correctly

Note that to restore the configuration, the backup must be fixed, too! In other words, existing cfg backups are incomplete in systems where nethserver-samba is not installed. The complete procedure is

1. install testing packages
2. run backup config

Then in another system

1. install testing packages
2. run restore config
   3 …

To update from testing run

 yum --enablerepo=nethserver-testing update

We’re fixing another bug affecting restore-config when network cards are different. @giacomo is investigating it! FIXED See NethServer/nethserver-base#75

[fasttech]

@davidep

Note that to restore the configuration, the backup must be fixed, too! In other words, existing cfg backups are incomplete in systems where nethserver-samba is not installed

Just to clarity, I have a populated snapshot, it’s an ad, with nextcloud, etc… therefore, I do not need to install testing packages into that before a backup… correct?

[davidep (Davide Principi)]

You need the fix before backup-config if nethserver-samba is not installed.

[fasttech]

@davidep

You need the fix before backup-config if nethserver-samba is not installed.

See… this was bothering me… now I know why, I’m not entirely stupid… just slow… I was thinking that ad dc was samba… but it’s nethserver-dc, and since I don’t have file server installed in that vm, nethserver-samba is not installed.

[giacomo (Giacomo Sanchietti)]

The bug is finally closed, maybe we will find some corner case in the future, but for now everything seem to work as expected.

To fix the issue we had to update 6 packages!

I would really say my thank you to @fasttech for this hunt and the hard (and boring testing)!
Big thank you also to @filippo_carletti for testing in a complex scenario, and of course thank you to @davidep who found a fix and implemented the working code!