Project

General

Profile

Bug #2192

app porting issues for f25

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

Status:
Assigned
Priority:
Normal
Assignee:
Category:
-
Target version:
-
Start date:
12/12/2016
Due date:
% Done:

0%


Description

let's keep track of app porting issues on f25 in this bug.

we can keep them all in one big bug and if some stick out we can make a separate bug for those.

i do se a lot of internal server errors during testing, but that may be because the system under test may have been wedged at that point.

History

#1 Updated by bigfoot65 almost 4 years ago

I just created a bug. I removed it since and will put the items here:

Noticing some issues with the testbot.

A. The MTU is rolling back to 1500. Have to manually set it to 1450 on reboot.
B. Many apps fail the test script with the error:

Unable to load URL: http://todoyu// because of error loading http://todoyu//: Service Unavailable

C. Some apps (Nagios) install packages that if removed take more than what was installed out.

#2 Updated by bigfoot65 almost 4 years ago

  • Assignee changed from bigfoot65 to cpg

Seems that PHP 7.0 has removed the php-mysql RPM a.k.a. PDO MySQL driver as known by apps. Most apps have not caught up to PHP 7, still using PHP 5.

As a result, this has broken many apps. In order to mitigate the problem, there is an RPM available from the REMI repo:

https://rpms.remirepo.net/fedora/25/remi/x86_64/php-pecl-mysql-1.0.0-0.11.20160428git45881bd.fc25.remi.7.0.x86_64.rpm

Recommend we add this to the Amahi repo. It can either be added to apps, or included on install of Amahi 10.

#3 Updated by bigfoot65 almost 4 years ago

Need to verify my last comment. It appeared to be the fix, but not 100% certain at this time.

The current php-mysqlnd module installed may be sufficient.

#4 Updated by bigfoot65 almost 4 years ago

Well looks like that was a dead end.

I am guessing that apps may not be compatible with PHP 7. Example is ImpressCMS. Works fine in F23, but not F25.

#5 Updated by bigfoot65 over 3 years ago

Noticed on the f25-bot that there are many entries in the /var/log/httpd/error_log file. Is there anything in there related to the issues with some apps not working?

#6 Updated by bigfoot65 over 3 years ago

Many articles I have seen on the internet state that PHP 7 dropped a lot of functions that were used in PHP 5.x.

This appears to be a contributing factor to many apps not working in F25. I tested Joomla and it confirmed this issue.

#7 Updated by cpg over 3 years ago

Aha. I see f25 has php-7.0.17.

we could try to install an older php, like in centos, with remi repos or something, but it can get dicy, i would imagine.

is there a signature to look for in the error logs or whatnot, so we can estimate how many apps are not enabled for php7 easily?

#8 Updated by bigfoot65 over 3 years ago

Yea I was installed PHP 5.6 from that repo on my test machine, but could not figure out how to force it to be used vs PHP 7.

That may be our best option to keep most of the apps working. The only indication of an issue in the log file lies in the webapp error_log. I would need to install each one, then check the log file.

A lot of work, but may be necessary.

#9 Updated by cpg over 3 years ago

i think this looks like a bit of a nightmare.

i think the best solution is to go fresh and only put out apps live that WORK on P7, and the rest let them port as time goes on.

delaying things further on our end will really not help much.

i will change apps that failed in the grinder to alpha, then we can try for a solid release, then look at apps one by one.

could you please add a visible WARNING in the wiki on the update from A9?
maybe even a warning in the A10 install page too. Not as prominent perhaps?

#10 Updated by bigfoot65 over 3 years ago

Agreed.

Do you have a goal date in mind for release?

I need to build the Amahi 10, Fedora 25 fresh install and upgrade pages. Will need to do screen shots and such.

As for apps, some of the failed apps work but fail due to java script, such as ownCloud and Feng Office. I will promote them LIVE as I test each one.

#11 Updated by cpg over 3 years ago

here is an alpha release of the platform with rails 5.x, which may or may not work well. se if you can give it a test on a clean f25 install.

upgrade to it, or you can simply install it instead of the usual hda-platform if you are testing from scratch, .. which would be a good test :-)

http://alpha.amahi.org/tmp/hda-platform-9.3.0-1.x86_64.rpm

#12 Updated by bigfoot65 over 3 years ago

Installed the test platform after installing hda-ctl. It installs the platform automatically, so the test RPM is an upgrade.

I received the following error when trying to access the dashboard:

The Dashboard has encountered an exception!

Woopsie. Sorry about that!

It could be due to a bug, or very low memory, or very low disk in the root (/) partition.

If your disk or memory are not critical, it's likely a bug.

If this persists, please file a bug report with as much detail as you can.

You will be redirected to a debug page with more information shortly.

For Troubleshooting visit here Troubleshoot

I also rebooted a second time with no change.

See https://da.gd/3mmQf for production.log

#13 Updated by cpg over 3 years ago

Thank you for the log. Bad cpg, no cookie. Not enough testing.

#14 Updated by cpg over 3 years ago

ok, one step at a time ... this one should install, and the front page should come up. i think the set up area may not work. apps tab definitely produces some exception i need to figure out.

no need to fetch the entire production log. or any of it for now, save you the trouble, as i am working on something related to the tab

http://alpha.amahi.org/tmp/hda-platform-9.3.1-1.x86_64.rpm

#15 Updated by bigfoot65 over 3 years ago

Tested with Fedora 25 DVD. The following are my observations:
  • Remove message from install:
    ***************        TBD: Initialize storage pooling ... *******************
    
  • Not using HDA for DNS message displays incorrectly:
    Your client device is probably not using your HDA for DNS. We recommend <a href='https://wiki.amahi.org/index.php/Transition_to_Amahi' target='_blank'>the following</a> for best results
    
  • The "not using DNS" message appears even when using HDA for DNS.
  • Amahi Initialization fails. After entering user name and password twice, the following error occurs:
    The Dashboard has encountered an exception!
    
    Woopsie. Sorry about that!
    
    It could be due to a bug, or very low memory, or very low disk in the root (/) partition.
    
    If your disk or memory are not critical, it's likely a bug.
    
    If this persists, please file a bug report with as much detail as you can.
    
    You will be redirected to a debug page with more information shortly.
    
    For Troubleshooting visit here Troubleshoot
    
  • After a few seconds, it redirects back to the Amahi Initialization page. Repeating the process yields the same error. Appears to be stuck in a loop. Cannot proceed from here.
  • Also verified rsyslog is installed, however it is the Fedora 24 version:
    Installed Packages
    rsyslog.x86_64                      8.12.0-3.fc24                      @anaconda
    

Standing by for further direction.

#16 Updated by bigfoot65 over 3 years ago

Will test with Fedora 25 CD too and report back.

Probably won't be until tomorrow.

#17 Updated by cpg over 3 years ago

ok, new rpm for testing. new user creation, tabs should work, networking, etc. but not apps or plugins

http://alpha.amahi.org/tmp/hda-platform-9.3.2-1.x86_64.rpm

#18 Updated by bigfoot65 over 3 years ago

Tested new platform with the following functions working:
  • Amahi Initialization
  • Add/delete user
  • Add/delete share
  • Advanced Settings
  • Log in/out for users
  • Network settings

The message about not using HDA for DNS is reflected even though the client was using the HDA as DNS server. DHCP was being handled by a different entity, so that may be why.

Also noticed the MariaDB server is still showing stopped (#1802). Some DNS aliases broken (#1342)

Good progress so far. I will say the DNS message that was added is a bit annoying. It's bright orange and red which shows on every screen all the time. Should be an option added to the Settings tab to turn it off. Users who don't use the HDA for DHCP/DNS will likely not want to see it.

#19 Updated by cpg over 3 years ago

the message will appear unless the client is accessing the hda as

http://hda

or hda.localdomain.com

in your case you do not say how you access it, so maybe it's just a matter of how it's written.

#20 Updated by bigfoot65 over 3 years ago

I was accessing via IP address but DNS was set to the HDA.

#21 Updated by cpg over 3 years ago

apps tab working, app install and uninstall should work, no more mariadb, apache in servers and disable dns removed:

http://alpha.amahi.org/tmp/hda-platform-9.3.4-1.x86_64.rpm

#22 Updated by cpg over 3 years ago

i think installing plugins may be broken due to some asset pipeline error in locating the plugin css file(s).
at least the disk wizard.

but i was able to install and uninstall two basic apps.

#23 Updated by bigfoot65 over 3 years ago

Awesome!

Testing an updating some apps now using the latest platform :)

If all works well, I will update the testbot as requested and do a mass test.

Think we are getting close.

FYI One thing I noticed is apps finish install in the log, but the UI seems to lag and sometimes does not ever finish. Going to Installed Apps shows it.

#24 Updated by bigfoot65 over 3 years ago

Installing a plugin has broken the new platform in my test VM.

Cannot manually uninstall the plugin using app-install script. It blows up :(

Not sure of the solution other than resorting to a known working VM snapshot.

#25 Updated by bigfoot65 over 3 years ago

The following apps failed the Grinder, but manual testing reflects they are working:
  • Chive: bad html
  • eyeOS: javascript
  • Nextcloud 11: javascript
  • OpenSIS: bad html
  • ownCloud 10: javascript
  • PHP-Fusion: unknown
  • Plex Media Server: bad html
  • Pydio: javascript
These plugin apps (currently LIVE) are pending testing once #2233 is complete:
  • Database Backups
  • Disk Stats
  • Disk Wizard
  • Greyhole UI
  • Web Apps

#26 Updated by cpg over 3 years ago

ok! this work good enough for a release!

with a pending work-around for #2233, which also requires code updates to the disk wizard (i think) as well as possibly other plugins ..

http://alpha.amahi.org/tmp/hda-platform-9.3.5-1.x86_64.rpm

i think if you could try the plugins (with the DW last) and report, it would be great!

#27 Updated by cpg over 3 years ago

i threw it on my dev vm and it seems like

  • disks stats works
  • greyhole ui works
  • web apps works
  • database backups works
  • disk wizard works! (probably because the app changes supercede that of the plugins)

i do notice it takes like 10 seconds and two full-throttled CPUs of ruby to bring up the dashboard after a plugin is installed, because the whole RoR stack is restarted.

#28 Updated by bigfoot65 over 3 years ago

Testing now and appears to be working as described.

I don't like the slow response after install though. Anything we can do about that?

Users might get impatient.

#29 Updated by cpg over 3 years ago

yes, we can disable the asset pipeline. we're looking into it. would you file a bug for that. ruby on rails is quite aggressive in pre-compilation.

it's after installation. and it depends on their hardware (most people have enough, but yeah, many will be slow), so it's a smaller concern, but agreed.

#30 Updated by bigfoot65 over 3 years ago

Ran plugin apps through Grinder with new platform installed. All passed!

Also available in: Atom