Devuan Jessie beta released

Posted May 2, 2016 19:04 UTC (Mon) by drag (guest, #31333)
In reply to: Devuan Jessie beta released by dlang
Parent article: Devuan Jessie beta released

> While I admit it's not as common, I am seeing Ubuntu show up more and more in the Enterprise.

This is true.

The wait from CentOS 6 to CentOS 7 was a big driver in this. Ubuntu was increasingly seeming as a viable alternative to Redhat because Redhat 'EL' release was getting so crusty and unpleasant to use.

RHEL 7 and CentOS 7 reverse this trend for my company. However we are still looking at Ubuntu for various reasons. Applications dictate the OS, not other way around. If the application says to use Ubuntu we are not going to argue with it.

on a side note:

For my personal setup I am re-re-re-learning how to setup Openstack in the 'correct' manner. As far as I can tell the easiest way to do this (which is generally the best) is by installing Ubuntu 14.04. And, holy shit, it was a culture shock. I have been using systemd-based distros for longer then most, since I use Fedora exclusively for my desktops, and going back to pre-systemd days was, to put it bluntly, very unpleasant.

One terrible bug that had me clawing at the walls was triggered by editing /etc/network/interfaces. I install Ubuntu 14.04 on my servers, but the installer defaults to DHCP. This is fine, but I want static IPs. I have a range of ips that are 'static' and range that are 'dynamic'. So I log into them, edit the network/interfaces file, and then ifdown/ifup and finished.

I go to bed, wake up, and later that afternoon I find that every single of one of my servers is now unaccessible.

Why? Because ifdown'ng the interface didn't kill the dhclient process. The Debian interfaces file has been around for, what?, 15 years? 20 years now? And it's still not mature enough to handle re-configuring eth0 correctly yet. This sort of stuff is a bit flabbergasting to run into.

This sort of stuff is why distro-specific management and configuration interfaces need to die. It's very hard to get right and if every distro does it's own thing then it needs to get done right in a hundred different ways by a hundred different projects. Chances of that happening is just about zero.

to post comments

Devuan Jessie beta released

Posted May 2, 2016 20:58 UTC (Mon) by rahvin (guest, #16953) [Link] (2 responses)

That's a weird error. Having only recently went to systemd with Jessie the network shouldn't have called the DHCP client after you edited the interfaces file. That could be something related to Ubuntu but the only time I ran into something like that was when the system was installed as a "desktop" and Ubuntu installed network manager, which takes over network management and still leaves the interface file there to screw everything up when you go to edit it. I broke networking so bad I gave up trying to fix it and just trashed the VM and reinstalled.

Devuan Jessie beta released

Posted May 2, 2016 21:26 UTC (Mon) by nybble41 (subscriber, #55106) [Link] (1 responses)

> Having only recently went to systemd with Jessie the network shouldn't have called the DHCP client after you edited the interfaces file.

The original configuration was with DHCP, and the interface was up, so dhclient was already running. After editing the interfaces file the configuration was static, so at that point ifdown didn't expect dhclient to be running and consequently failed to terminate it.

The problem is modifying the configuration file while the interface is up, combined with scripts that use the *current* interfaces file for ifdown rather than the configuration which was in place when ifup was executed.

Under systemd any background processes like dhclient associated with an interface (in the corresponding cgroup) would automatically be terminated when the interface is stopped, regardless of the current configuration. This is not a perfect solution to the problem of changing configurations, but it does mitigate the specific issue of ifdown failing to stop dhclient.

Devuan Jessie beta released

Posted May 3, 2016 19:23 UTC (Tue) by rahvin (guest, #16953) [Link]

Now I understand what happened. Thank you. These are the type of errors that make you want to tear your hair out and make me quite glad that sysv init is being replaced by a sensible program that actually monitors dependencies in daemons.

Devuan Jessie beta released

Posted May 2, 2016 22:52 UTC (Mon) by dlang (guest, #313) [Link] (1 responses)

>> While I admit it's not as common, I am seeing Ubuntu show up more and more in the Enterprise.

> This is true.

> The wait from CentOS 6 to CentOS 7 was a big driver in this. Ubuntu was increasingly seeming as a viable alternative to Redhat because Redhat 'EL' release was getting so crusty and unpleasant to use.

Actually, where I see Ubuntu making the biggest penetration isn't where there is an existing RHEL setup, but rather in the smaller companies just introducing Linux, or moving from 'joe set something up' to a more format setup. At that point there isn't an installed base to compete against, and so the choices are based on what people know.

How distros get picked

Posted May 3, 2016 8:22 UTC (Tue) by tialaramex (subscriber, #21167) [Link]

That's roughly why there's RHEL on all our gear today, certainly. We had Fedora boxes when we first started, to get (by Enterprise standards) bleeding edge features, at that point some of the hardware was physically in someone's home. Then we had CentOS machines, by now everything was in an actual data centre in racks and we were hiring sysadmins. When we were purchased by a huge corporation obviously it needed to be an Enterprise offering that their teams of operations people would run for us, somehow, so RHEL. But it all starts with two key people running Fedora on their own machines before the business even existed.

Anyway, companies doing this stage will tend to be smaller, and the choices will get made by technical or semi-technical people, rather than by suits who need to see words like "enterprise" and "commitment" and want to feel as though they can phone up somebody who also wears a suit and yell at them if it breaks, as if that'll actually help. Also of course in 2016 the choice will often actually be "cloud", or at least leaving the hardware problems to somebody else, whether with the actual "cloud" or some managed private servers.