Koozali.org: home of the SME Server
Contribs.org Forums => Koozali SME Server 10.x => Topic started by: Fumetto on July 04, 2021, 01:49:31 PM
-
I made my first upgrade from version 9 to version 10. Same hardware, two new HDDs to be able to go back in case of problems. Console backup on SDD with USB adapter. Backup that I had to do twice as I initially formatted the SSD disk in ext3 with Windows and I was unable to import the backup due to a file system incompatibility; I put back the version 9 disks, formatted the SSD disk from the SME9 console, and first problem solved. :-)
For the installation of the system I proceeded to create a USB stick with Rufus and the ISO of SME 10. Everything ok, but I had to use the installation in text-mode because in graphic mode I was not able to create the Raid1; probably my incompetence, but in the end everything went well. There remains the console with the English keyboard to fix (I have an Italian keyboard) but I have seen that there is another tread (https://forums.contribs.org/index.php/topic,54527.0.html) about it and I will take a look at it.
Server with two network cards, server and gateway mode; about 30GB of data, server used only as a mailserver with about 2 / 3GB of data on a single user folder (no ibay). Everything imported correctly. After the reset and a couple of reboots I had to reconfigure the network cards (the only active interface after the reset was the loopback interface) but during the reconfiguration it also correctly imported the pre-assigned IP addresses on SME9.
The sequence used was: basic system installation, data recovery from backup, network interface reconfiguration and subsequent system update (yum update).
The backup to workstation on NFS share residing on a NAS was successfully performed, it also imported this configuration correctly.
SME9 only had sme9admin installed as contribs; installing smeadmin for version 10 did not import the data cows, but I expected it; never mind, I only use it to monitor temperatures and quantity of emails sent / received / discarded. However, it correctly imported the previous configuration (temperature sensors, ip address for ping statistics, amount of daily notifications, etc.). I also installed and used "qmHandle mail queue manager" and everything worked as expected.
The server certificate has been regenerated (I remembered that in the transition from SME8 to SME9 he had instead imported the old one, but I probably don't remember well, old age advances and memory is scarce).
"Logwatch" was installed automatically (I found the daily messages in the admin emails and I have to study it, it was a (nice) surprise).
Apart from some cursing to $_ancestral_gods due to the fact that the "new" HDDs were actually a recovery of an old raid installation of OpenMediaVault (and so I had to "reset" them after realizing it, a couple of hours lost ) all went well.
An infinite thanks to all who have worked to make this possible !!!
I will try, I think within a couple of weeks, to upgrade a VM on ESXi. I will therefore try to do a clean installation and a subsequent import of a complete backup on workstation made before turning off the SME9, again from NFS share on NAS. However, this VM has about 300GB of email-only data, no i-bay files or user folders. So it will take longer. From the release notes I understand that it is a supported procedure; is this correct? Any suggestion regarding this procedure, given the amount of data and the amount of time that the upgrade will require, is welcome !!!
And again thanks to those who have allowed all this !!!
-
Always good for the knowledge base to see others experiences and hiccups :-) and a restore using a workstation backlup rather than the console, good to hear this option is working as expected, good job.
Have fun and enjoy..
-
raid 1 is automatic with two similar disks, unless (as mentioned everywhere except in the efi boot menu) you enter in the menu to configure disk and then you loose it immediatly as you get the automatic default of centos which is boot partition then all available space under lvm.
you can still manually configure a raid 1 using both gui, your own kickstart, and text but you need to learn how.
smeadmin is supposed to import the data from sme9admin. at least it does for me. but you need to restore it first. default backup does not include /var/lib/sme9admin
regarding certificate, yes it is upgraded because needing a stronger key 4096. and now includes all ip and server names / domains.
for your vm, i suggest a dedicated virtual disk for the users emails so you can migrate easily next time.
-
OK, so my servers are now all running SME 10. Mailserver, Fileserver, Mediabox for backups, SME10 dev VM. Some on hardware directly, some on VMWare ESXI. The sticking points have been:
- the network device usually needed to be re-set by going to the admin manager and stepping through and selecting the right network card again.
- The installer is touchy - if you don't click exactly right, or pause too long, or click too fast, it has a hissy fit and crashes.
- PHP versions - some of the systems we use are based on really really old versions of PHP and didn't translate well. It's a relief that our servers are now getting updates again!
The first server I tried to setup the drives in another box first, and then just transfer the drives into the actual server. However, that didn't work for some reason - it showed a bunch of error messages and just gave up. For the rest of my upgrades, I would shutdown the hardware, put new drives in, run through the install/restore process.
For one I had to delete the drive partitions first because the installer would complain with a "Value error". I removed the LVM setup using fdisk, then the install proceeded smoothly.
After seeing only a fraction of what's gone on to make SME 10 a reality - big kudos to all those who have worked so tirelessly - thank you!
-
regarding certificate, yes it is upgraded because needing a stronger key 4096. and now includes all ip and server names / domains.
Does it do it for ALL domains? I have a number of domains but they are all handled by Letsencrypt and not required in the selfsigned cert?
-
yes for all configured on the server.
idea is to be ready to handle situation where external certificate becomes bogus.
it is then used as a failsafe.
-
Does it regenerate when a domain is added or removed?
-
yes.
-
Kk.
That'll take some time on some boxes I'd guess!
-
Kk.
That'll take some time on some boxes I'd guess!
while LE / dehydrated takes some time to validate all the domains one by one, the self signed certificate is pretty fast to generate with all domains, hostnames, hosts and IPs.
-
Second upgrade made: it was a VM with SME 9 updated up to the date of EOL and only two contribs installed, sme9admin and qmhandle.
Backed up on workstation via NFS, off old VM, installed new VM, set FQDN and IP as old VM, scheduled backup on workstation (install nfs-utils), restore from backup previously made via server manager. Reinstall the two contribs. Everything is OK.
A couple of notes: the server manager, in the restore page, times out (and with 300GB of data I was almost certain), I had to "check" from the console that the DAR process had finished in order to restart (post-upgrade & reboot), network reconfiguration from console and ... everything seems to work !!!
He also "got" the old DKIM configuration for the mail sent (which I was in doubt he would bring with him) ... now I have 24 hours to test everything before putting the server back online ready for 16/08.
I just have to study to install support for vmware tools ...
-
please report as bug the timeout for restore.
-
please report as bug the timeout for restore.
Done (https://bugs.koozali.org/show_bug.cgi?id=11670), but with 200GB of restore can be "normal" a webpage timeout...
-
or we refresh the page like yum regularly and take care to renew the CRSF token which is not yet handled in the yum page
-
please report as bug the timeout for restore.
There is still a sme9 bugs re this, https://bugs.koozali.org/show_bug.cgi?id=8271
Long standing..