On Monday, December 16, 2019 12:52:49 PM MST Kamil Paral wrote: > On Mon, Dec 16, 2019 at 6:14 PM John M. Harris Jr <johnmh@xxxxxxxxxxxxx> > > wrote: > > On Monday, December 16, 2019 9:56:01 AM MST Adam Williamson wrote: > > > This is not accurate. You're not accounting for the time it takes to > > > write the disc, and we also have to check that the media check works, > > > which takes quite a while on its own. > > > > I was accounting for that time. Writing to a disk and checking it does not > > take a long time, ~20 minutes at most. This is why I said "less than an > > hour" > > in total. Additionally, during that time, no user interaction is required, > > once the process is started. > > I guess I should provide some better data for those estimated time > requirements. I haven't used a stopwatch during the last release cycle, but > my estimate is that it takes 1-1.5 hours to check a single install medium. > This includes burning the DVD, booting it in BIOS mode including the > mandatory and default media check, performing the installation, and then > repeating the boot and install in UEFI mode. Occasionally there are some > optical reading-related issues, e.g. when a machine gets stuck because it > constantly spins up and spins down the disc, having a problem trying to > read some area. Sometimes the disc access gets unusably slow, just to work > fine after a reboot. All the usual stuff that you come across when using > CDs/DVDs. Some of that is definitely caused by our rewritable media being > scratched, or DVD drives being old and the laser no longer being well > calibrated. We'd have to buy new drives to improve that experience, but I > don't really see much sense in that, when optical media is a niche > technology nowadays (hence this proposal). It is simply not the case that optical media is niche, especially not in enterprise installations, low end consumer, business class systems or new old stock systems. For example, at $WORK, we use optical media for all > We have 2 release-blocking media, so the total time is somewhere between > 2-3 hours (likely closer to 2 hours, because netinst installation is way > faster due to downloading packages from the net instead of copying them > from the disc). That's not the main problem, though. The main problem is > that during that time, one or two of our test machines in our office is > fully occupied with spinning the discs, and we can't use it for anything > else. That means all other bare-metal testing needs to wait. As Adam > already pointed out, sometimes we need to check the final candidate > composes in a single day, i.e. in the standard 8 working hours (and yes, we > often work overtime in these cases). Blocking half of our bare-metal office > test machines for 2 hours out of 8 is not a small deal. Do you need more test hardware? Honestly, that's what this sounds like. > It's simple to say "no user interaction is required", but that's not > completely true either. If you want to do the QA job properly, you need to > have an eye on the media consistency check, because we've had issues in the > past where it timed out and either considered it a pass or fail (both are > incorrect). So you can't simply walk away and come back and consider it OK > when it reached the installer, you really need to watch the progress in > certain critical points. Once the UI is ready, it is much slower than when > booting from USB. So you often spend 10, 20 seconds staring at the screen > until it decides to do something. Is that due to the hardware under test, or is it a result of scratched media? > The actual installation progress is unattended yet. But you need to check it > frequently to see whether it finished, so that you don't waste time of the > bare metal machine standing idle. There are many more tests waiting in the > queue. > The fact that this whole process is a major annoyance (it really makes you > hate optical media, if you deal with this regularly) is of course > contributing to the fact that we don't want to do it anymore. We're only > humans. But we wouldn't have proposed the criterion change if we hadn't > thought the time is right and that it is no longer an important factor for > the majority of our users. We've waited very long with this proposal. And I > still intend to keep testing optical media functionality from time to time, > even when optical-blocking criterion is removed. But I'll do it once or > twice per cycle, probably with a Beta GO compose, and not for every release > candidate created. I'd happily volunteer to help test this, but this is the standard method of installing in many environments, and is also the ONLY option in a good number of environments as we've described. -- John M. Harris, Jr. Splentity _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
