test-analytics

License

download KVM image of AppScale-ready VM http://download.appscale.com/apps/AppScale%201.13.0%20KVM%20Image
tar xvzf appscale-1.13.0-kvm.tar.gz and go enjoy a cup of tea/coffee
- do not bother using your pc for anything else while this is running (YMMV, depends on how bad your IO scheduler is)
import the .img file as a new VM in virt-manager or whatever, at least 3gb ram recommended
start the VM
login as root with password appscale
run appscale init cluster
modify cluster config file vi AppScalefile
- change the IPs in ips_layout to the VM IP
- to make your life easier, consider adding these lines:
```
admin_user: root@localho.st
admin_pass: 123456
```
appscale up, enter admin e-mail and password (eg. root@localho.st and 123456) when/if prompted
you should now have an appscale cluster up and running
git clone https://github.com/saironiq/test-analytics.git
cd test-analytics/testanalytics_frontend && mvn gae:unpack && mvn gae:run + another cup of your favorite hot drink
press ^C after the dev servers starts up (no maven target to build the package o.O)
mv target/{test-analytics-1.0-SNAPSHOT,war} to make appscale happy
cd && appscale deploy test-analytics/testanalytics_frontend/target/ and enter some e-mail address again (no idea what it's good for)
after the app starts up successfully appscale relocate test-analytics 80 443 to make the app available on standard ports
???
profit! (or not, continue reading below)

appscale 1.14.0 does NOT work, so do not bother trying... (at least the KVM image)
- hangs at waiting for dashboard app to come up or smth like that
cluster does not autostart after reboot BY DESIGN - you have to manually log into the head node and do appscale down && appscale up (cause apparently something is running o.O)
- irc convo log with some relevant info: http://meetbot.eucalyptus.com/channel-logs/freenode/%23appscale/%23appscale.2013-06-24.log
app and its data SHOULD persist across reboots (fingers crossed)
you need to manually relocate the app AGAIN if you changed the ports it should be listening on (dafuq?! >.<)
SSL certs are different after each cluster start-up (minor issue compared to the ones listed above *grumblegrumble*)

TL;DR: appscale sucks

Ubuntu is an ancient african word, meaning 'I can't configure Debian.'

~ definition of Ubuntu from urbandictionary.com

out-of-date python (2.7 required for appscale, 2.6 in rhel6)
requires external repositories/specfiles for multiple packages (eg. rabbitmq, cassandra, zookeeper)
incompatible version of monit in repos, had to be built from source
almost every package has a version different from the corresponding ubuntu pkg - possibly incompatible versions
various smaller "bugs" not present in ubuntu like missing files and directories during cluster start-up (possibly due to failing python scripts)
rake unit tests for appscale failing due to outdated python
when attempting to start appscale cluster anyway, the appcontroller can't run/reach the cassandra db (it should start it automatically but doesn't and even when it's started manually nothing happens)
all the cluster core services are getting killed wihout being started again (nginx, memcached, monit, ...)
the whole thing just doesn't work -.-"

missing erlang, mock, django, ejabberd, rabbitmq, python-pip, etc...
incompatible ruby (1.8 required, has 2.0) - dependencies fail to build ("rcov" gem)
- this is a serious problem as the whole appcontroller server is written in ruby