War Story: distro version upgrades can be dangerous

One of my gateway machines uses Fedora. I used to use CentOS for that kind of purpose, but no longer. When CentOS changed, I decided to try Fedora instead. The first downside is the torrent of updates. But that doesn't really seem to be a problem. The second downside is that support for a release of Fedora is two and a bit 6-month release cycles. An upside is that version upgrades are automated (unlike CentOS) and that automation generally works. This upgrade takes the machine offline for perhaps an hour at a time of your choosing (If one isn't diligent, one can fall way behind upgrading distro versions. This is particularly easy in systems like CentOS or debian where there is a combination of long support periods and no automated version update procedure. Ubuntu LTS appears to get this right: long support and update automation.) Fedora has been working pretty well on my gateway for the last few years. I like it that the packages are up to date (CentOS packages tend to be old). I upgraded the gateway machine from Fedora 36 to 38 a couple of days ago. It didn't go smoothly. A couple of small-ish hangups caught me. This is not ideal in a gateway. Luckily, I have more than one. ==== Problem 1: firewalld firewalld had a new bug. Or maybe it was iptables. I don't need to know which. The firewalld settings could not be loaded. So the firewall had default settings. These were safe but not functional for a gateway machine. I could manually install what I wanted in the running firewall but those settings would not persist. I could not make persistent changes. I reported the bug: https://bugzilla.redhat.com/show_bug.cgi?id=2209199 Within a day or so, a work-around is suggested and a bug fix is initiated. This shows outstanding support for which I pay nothing. (Actually, I have done free support for specifically for Red Hat as part of an upstream project. But there is no connection.) ==== Problem 2: GNOME now suspends the machine when it thinks nobody is using the console. See <https://discussion.fedoraproject.org/t/gnome-suspends-after-15-minutes-of-user-inactivity-even-on-ac-power/79801> Guess how I found out? No, not reading the Fedora 38 release notes. I discovered it when my gateway suspended. What the heck?!? This is a disaster for any machine that has tasks that need to run all the time. Clearly this applies to a gateway machine, but even on my desktop machine, this isn't correct: - I want to be able to SSH into my desktop at any time - I run Hexchat IRC client 24/7 on my desktop so that I can know of activity while I'm not there. - My desktop is an internal mail server and needs to handle mail at all times. There is a "Fedora Server" version that gets this right. I haven't figured out how to get Fedora Workstation to stop suspending generally. The work-around is that you can get each possible console user to turn this feature off. And you have to do this for GDM (the login screen) too.

| From: D. Hugh Redelmeier via talk <talk@gtalug.org> | I upgraded the gateway machine from Fedora 36 to 38 a couple of days ago. | | It didn't go smoothly. A couple of small-ish hangups caught me. This | is not ideal in a gateway. Luckily, I have more than one. I recently noticed another third problem: BIND / named wasn't brought up. This was all in the system log but I didn't read it. The problem was that for some reason SELinux policy wasn't updated so SELinux prevented BIND from accessing /var/named See https://bugzilla.redhat.com/show_bug.cgi?id=2211935 SELinux is kind of complex and tricky but the error messages are clear and helpful. This seems to be a rare problem but luckily someone else hit it before I did.
participants (1)
-
D. Hugh Redelmeier