Fog Installer - Distro check
-
The failures are infact a problem with Fedora. See this thread for details: https://forums.fogproject.org/topic/11702/upgrading-from-1-5-0-to-1-5-1-fails-at-backup
-
As of now, all installer tests are passing with the release of 1.5.2.
-
I’m going to try to get Ubuntu 18.04 into the test cycle tonight.
-
@wayne-workman Do you have a way to do a quality of service check against FOG or are you doing this right now?
What I’m thinking is a http curl call to the vm at the www root and see if the url was redirected to /fog/management or something like that . That would test if the fog UI was actually up and talking with the database.
-
@george1421 I could add this in - but as soon as this is in there, we will want to check something else next. There is a project called Selenium that is suited for testing responses to web requests. @developers @moderators does anyone have experience with Selenium that could help put together some tests? If not, I’m fine with writing up a simple python script to do some testing with the Requests library (I have experience with that).
-
@Wayne-Workman Back in 2007 I used twill when doing Web Application Firewall testing for my diploma thesis. But seems like this project hasn’t gone anywhere since then. Just thought I might throw that in, probably not worth anything.
-
@sebastian-roth It’s truly dis-heartening to me to be shown awesome libraries from yesteryear that didn’t take off… My heart goes out to Twill. However, Requests has gained substantial community support (not due to any of my doings) and probably has a much longer future than other libraries. I was hoping someone had Selinium experience - but I really do think we can do just fine with Requests.
-
A heads up - seems like Ubuntu 18 is having a lot of problems these past few days. I’ve not looked into it yet.
-
I’ve added Fedora 28 server to the list.
-
Looks like there is a possible issue with php-fpm on Ubuntu 16. I’ll look into it in more detail this evening.
-
@developers I’ve looked into the Ubuntu 16.04 problems for the last two hours and have determined the problem is being intermittently caused by a problematic ppa.launchpad.net server:
91.189.95.83
. Theapt-get update
command times out, and because it was not successfully run, lots of packages fail to install. I’ve also looked over all of the commits in the working, dev-branch, and master branch of the fogproject github repository around the days that the Ubuntu 16.04 failures started happening, there are no commits that would cause this.The workaround: If the installer fails for you, manually run
apt-get update
until it succeeds. Once it has succeeded, run the installer again and it should work. -
@Wayne-Workman Thanks heaps for digging into this!
-
Just an update,
You will notice that the link in my signature is different, the one there is the new one.
Long story short, I setup an email server for myself, which caused DNS & ssl problems to arrise with the fogtesting dashboard. So - I moved the site to a t2.nano in AWS. The same VM here at my house is still doing 100% of all testing, but when it gets done it does a quickrsync
to the aws box to update the website with the new dashboard & files. -
I’ve added Fedora 29 Beta to the daily installation tests. I’ll remove it and add the official release as soon as that’s available. I’ve also updated all of my ‘clean’ snapshots with the latest patches.
-
I’ve removed Fedora29-Beta, and setup a Fedora29 minimal server today, from the official release ISO. I’m pretty sure I remembered to do everything this time (like firewalld and selinux). We’ll see how it goes tomorrow morning.
-
So the issue with Fedora 29 is the Remi repo, they don’t seem to be caught up with the recent release. Their website still lists the beta version. I tried manually installing the remi repo for fedora 29 beta, doesn’t work. I don’t think there’s anything we can do until Remi Repo gets their end fixed. Most likely, when it’s fixed, everything will just start working because it appears the fog installer is creating the correct commands.
-
So as expected, the remi repo removed Fedora 29 beta, and added Fedora 29 release. Dev-branch and master are passing the installation tests, but working is failing which seems strange. Ubuntu 18 on working branch also failed - this test was run immediately after the Fedora 29 working test, so maybe it’s something with my vm host or internet at that time. It’s normal for this test system to have the occasional blip (it is what it is) so we’ll see what the tests say tomorrow.
-
@Wayne-Workman About Fedora 29: I do see the following message over and over again:
Remi's RPM repository - Fedora 29 - x86_64 47 B/s | 2.3 kB 00:49 Failed to synchronize cache for repo 'remi', ignoring this repo.
Not exactly sure if that is causing the broken installation.
Now Ubuntu 18 on the working branch. I really have no idea what is going wrong here. It seems to stop right in the middle of an apt… call. From the last lines of log output it looks like it. Possibly your test install VM got interrupted!!!
-
@Sebastian-Roth Yesterday’s issues were just a blip. All green today.
-
Seems Fedora29 is having some problem related to apache.
These are in the logs:[Thu Nov 29 05:20:30.553661 2018] [proxy_fcgi:error] [pid 8980:tid 140162887972608] [client 10.0.0.48:40096] AH01067: Failed to read FastCGI header [Thu Nov 29 05:20:30.553706 2018] [proxy_fcgi:error] [pid 8980:tid 140162887972608] (104)Connection reset by peer: [client 10.0.0.48:40096] AH01075: Error dispatching request to :
Also - these logs and the dashboard are in AWS s3 now. The bash scripts that makes this stuff is adding an extra
/
somewhere and s3 does not handle this like Apache. So when you click the links and it says “page not found” you just need to remove the extra slash.