Project

General

Profile

Bug #1845

Fedora 23 report

Added by cpg almost 4 years ago. Updated over 3 years ago.

Status:
Feedback
Priority:
Normal
Assignee:
Target version:
-
Start date:
09/26/2015
Due date:
% Done:

0%


Description

so far things look to be working relatively ok on the latest publicly released fedora 23 install.

  • dashboard up and running
  • no blog news, no "all new apps" image
  • there is no external dns working. something is up with this. changed to google public dns and it worked switched to others, stopped
  • hda-install needs some fixes
  • the machine needs a little patching
  • we need a new repo for f23
  • mariadb is running but the server light is red (known issue)
  • monit not quite well, though it did not crash
  • missing rsyslog
  • no apps tested

History

#1 Updated by cpg almost 4 years ago

  • Description updated (diff)

#2 Updated by cpg almost 4 years ago

  • Description updated (diff)
  • Status changed from New to Assigned
  • Assignee set to cpg

#3 Updated by bigfoot65 almost 4 years ago

App installs fail:

=======  app install begin      @  2015-09-26 20:08:05 -0500 ==========
Installing   app id 2c50o3i8x1 under /var/hda/platform/html/script/.. ENV=production
/var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/connection.rb:144:in `handle_response': Failed.  Response code = 404.  Response message = Not Found. (ActiveResource::ResourceNotFound)
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/connection.rb:123:in `request'
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/connection.rb:82:in `block in get'
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/connection.rb:216:in `with_auth'
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/connection.rb:82:in `get'
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/base.rb:990:in `find_single'
        from /var/hda/platform/html/vendor/bundle/ruby/gems/activeresource-4.0.0/lib/active_resource/base.rb:877:in `find'
        from /var/hda/platform/html/app/models/app.rb:187:in `install_bg'
        from /var/hda/platform/html/script/install-app:86:in `install_or_uninstall'
        from /var/hda/platform/html/script/install-app:107:in `main'
        from /var/hda/platform/html/script/install-app:119:in `<main>'

Tried WordPress and Adminer with no luck.

#4 Updated by cpg almost 4 years ago

  • Description updated (diff)

#5 Updated by bigfoot65 almost 4 years ago

After the changes you made, tested app install/uninstall for Adminer. Worked perfectly. Also installed phpSysInfo and it worked as well.

#6 Updated by bigfoot65 almost 4 years ago

Switched DNS to use OpenDNS and appears to be working correctly. Will monitor for any further problems.

#7 Updated by cpg almost 4 years ago

After updating the f23 repo with fresh builds of hda-ctl and hda-platform, i'm attempting to upgrade my F21 HDA to F23:

# dnf update --refresh
..
# dnf install dnf-plugin-system-upgrade
..
# dnf system-upgrade download --refresh --releasever=23
...
2596 packages downloading ...
...
# dnf system-upgrade reboot

#8 Updated by cpg almost 4 years ago

ok, i left the system to its own devices at that point. i deliberately not accessed the console.

after a while (20 minutes or so), i still saw the activity light and eventually gave up because i saw it active, but the network was not coming up.

connecting the console, i saw the following:

  • the network was up, with a dynamic ip. type this to find out: ip addr
  • network device had been renamed (OUCH) from ps1ps2 or some such to enp2s0. This will give us endless headaches. I had to do as follows
    cd /etc/sysconfig/network-scripts; mv ifcfg-p1p1 ifcfg-enp2s0
  • then i had to edit the file to make the device name match. the rest of the settings would work well.
  • I then typed hda-ctl-hup because the /etc/resolv.conf file had bad settings, so I figure maybe the hda-ctl-hup would re-sync the settings up
  • finally, the network came up, however, no devices were getting IPs vis DHCP. the firewall was set to enabled and was turned ON!. So opinionated! To disable:
    systemctl disable firewalld; systemctl stop firewalld
  • after this, things started to work more or less
  • the dashboard was not working with an exception. The file /etc/issue has no information about the version of the OS, so we should use /etc/system-release. both hda-platform and hda-ctl need changes for this.
  • some core services like samba, mariadb, httpd and dnsmasq appear to be running healthily
  • some apps seem to work without even uninstalling and installing them (transmission, sickbeard, sabnzb+), but not amahi time machine
  • Amahi Anywhere came up stopped and it shows in status now as "vendor preset: disabled" though it's enabled.
  • Amahi Sync seemed to come up running, and enable, but also vendor preset: disabled
  • The amahi-greyhole service was not running, but it was also enabled and v-p: disabled

I do notice that the fedora cockpit is by default running at https://hda:9090/ with a self-signed cert. Not sure if exposing this by default is a great idea (giving users enough rope to hang themselves and all that). We may add an alias and a virtualhost for this. Maybe.

#9 Updated by bigfoot65 almost 4 years ago

Performed an install from scratch of F23 using CD with minimal install option. Did not set root password, but made user administrator.

Fedora installed fine. Began Amahi 9 installation and all went well until the hda-install CODE portion.

Upon execution, this is the result:

[root@AP-F23-Server-CD ~]# hda-install TPKHBHS
Verbose log file at '/root/hda-install-151109-201400.log'.
Retrieving install code ...
Install code looks good
Starting MySQL
execution of '/etc/init.d/mysql start' failed with code pid 1034 exit 127

Install ended at that point.

#10 Updated by cpg almost 4 years ago

There is a fix for hda-install that will be pushed out soon.

however after using F23 for a few days, i notice the following behaviors:

  • the firewalld service keeps on being turned on at boot time. this messes up DHCP service at the core, but pretty much everything else
  • amahi-sync starts, etc. but not amahi-anywhere. i suspect some dependency in the systemd definition or something
  • auditd keeps on filling the messages file with lots of garbage

#11 Updated by bigfoot65 almost 4 years ago

Did a test install of Fedora 23 minimal install CD and Amahi 9. All worked as expected and no errors.

However, when accessing the dashboard, I get the Apache Test page.

#12 Updated by cpg almost 4 years ago

great, some progress. thanks for testing.

i think just removing the /etc/httpd/conf.d/welcome.conf was enough to get things going.

otherwise, if i can access the vm, i could debug more.

#13 Updated by bigfoot65 almost 4 years ago

Deleted the file but did not help.

You can access it via SSH. I will email you the credentials.

#14 Updated by cpg almost 4 years ago

ok, i did the following:

  • remove all the userdir, welcome and autoindex conf files from /etc/httpd/conf.d/
  • tried to check status with # passenger-status
  • restarted httpd, etc.

strangely, it worked, though passenger seems to show it's not running (though i see processes from Passenger*):

# passenger-status
ERROR: Phusion Passenger doesn't seem to be running.

Networking

so, i have a question -- the dhclient program was running, i noticed.
did you set your DHCP server to give this VM the proper expected IP?

i set up the ifcfg file for the device to be static on the same IP it had (.23).
we will have to figure this out, since it was not static. same thing happened on my hda (which was an upgrade and the device name changed)

#15 Updated by cpg almost 4 years ago

Don't forget to answer the questions above. In the mean time, a new version of hda-ctl was release that removes {autoindex,welcome,userdir}.conf files from /etc/httpd/conf.d/ so that it does not interfere with the platform.

new installs hopefully will work out of the box after the final reboot, i'm hoping. the troubling part is that i am not sure.

#16 Updated by bigfoot65 almost 4 years ago

did you set your DHCP server to give this VM the proper expected IP?

Yes I did, but since have removed it.

I will test a new install once we get the testbot moved to OS. Didn't want to lose the work you had done.

#17 Updated by cpg almost 4 years ago

it's no problem. all the changes were committed to version control, so it's not hard to get it going.

#18 Updated by bigfoot65 almost 4 years ago

Did a new install via the CD. Installed Amahi 9 and all went well. The dashboard is displayed and works as designed thus far.

One problem is the static IP address is not getting assigned. I removed the Fixed IP from my network before install and the HDA now is getting a DHCP lease vs the static.

#19 Updated by cpg almost 4 years ago

the f23 rpmfusion packages are in our repos. however, note that i do not see any packages in their repos at this time http://download1.rpmfusion.org/free/fedora/releases/
unless of course the url for their repo for f23 has changed and it's not that one.
so, the rpmfusion "apps" may look like they work, however.

also, my updated system picked up the rpmfusion rpms already somehow. i am not sure how. maybe they make them autoupdate or something.

#20 Updated by bigfoot65 almost 4 years ago

This is a bug cpg filed for the firewall not being disabled on reboot even though it was manually disabled. It appears to only affect machines updated from Fedora 21 thus far.

https://bugzilla.redhat.com/show_bug.cgi?id=1288758

#21 Updated by cpg almost 4 years ago

looks like the rolekit service requires the firewall and that forces it to be run.
we will have to see what to do about this, probably disable rolekit, which is what i am doing for now. testing how it works long term.

#22 Updated by cpg over 3 years ago

  • Assignee changed from cpg to bigfoot65

we have pulled two updates to the f23 repo:

  1. make rolekit disabled, which was forcing the firewall enabled
  2. a difficult api change to interact with NetworkManager (what did you expect - something easy related to NM!??)

with this, if these work in a fresh install, i think we'd be very close to a beta release.

if you can test a fresh install, it would be great.

#23 Updated by cpg over 3 years ago

i am testing a fresh install and i found one issue: hda-ctl needs to stop quicker, rather than wait until it's new wake cycle, otherwise, it leads to a long wait on a shutdown:

https://dl.dropboxusercontent.com/u/364883/Screenshots/aeiu.png

#24 Updated by cpg over 3 years ago

as a CYA option, i added an alternative way to configure the network via the nmcli command to control NetworkManager.
it's invoked if any of the API calls to NM fail.

#25 Updated by bigfoot65 over 3 years ago

  • Status changed from Assigned to Feedback
  • Assignee changed from bigfoot65 to cpg

Did a test install and all went well. Once Amahi was installed following wiki guidance, it assigned the static IP address and the dashboard was available.

Everything appears to be working, but I will need to run the platform checklist to confirm.

#26 Updated by bigfoot65 over 3 years ago

Performed an upgrade from Amahi 8 to Amahi 9. There were no apps or Greyhole installed.

All went well and no issues were captured.

Will test again, but with some apps installed.

#27 Updated by bigfoot65 over 3 years ago

Did an upgrade from Amahi 8 to Amahi 9 this time with apps installed. They were as follows: Disk Wizard, Disk Stats, MediaWiki, SABnzbd Plus, Webmin, phpMyAdmin, DLNA, RPM Fusion (Non-Free), and RPM Fusion (Free).

All went well and no issues were captured.

During the upgrade, there were prompts to accept the GPG key on 3 occasions. This was for some RPMs from the RPM Fusion and Fedora 23 Repos.

The best option may be to skip the GPG check in the upgrade instructions to keep things simple:

fedup --network 23 --nogpgpcheck

A quick check of functionality reflects all is working well to include installing an app (phpSysInfo). A more detailed verification is needed.

Also available in: Atom