modulesync 2.12.0 / zabbix::web: Allow httpd to speak to the database #680
Conversation
|
so:
diff --git a/spec/acceptance/zabbix_application_spec.rb b/spec/acceptance/zabbix_application_spec.rb
index 61129f8..b6d6d65 100644
--- a/spec/acceptance/zabbix_application_spec.rb
+++ b/spec/acceptance/zabbix_application_spec.rb
@@ -38,7 +38,15 @@ describe 'zabbix_application type' do
# setup zabbix. Apache module isn't idempotent and requires a second run
it 'works with no error on the first apply' do
# Cleanup old database
- shell('/opt/puppetlabs/bin/puppet resource service zabbix-server ensure=stopped; /opt/puppetlabs/bin/puppet resource package zabbix-server-pgsql ensure=purged; rm -f /etc/zabbix/.*done; su - postgres -c "psql -c \'drop database if exists zabbix_server;\'"')
+ cleanup_script = <<-SHELL
+ /opt/puppetlabs/bin/puppet resource service zabbix-server ensure=stopped
+ /opt/puppetlabs/bin/puppet resource package zabbix-server-pgsql ensure=purged
+ rm -f /etc/zabbix/.*done
+ if id postgres > /dev/null 2>&1; then
+ su - postgres -c "psql -c 'drop database if exists zabbix_server;'"
+ fi
+ SHELL
+ shell(cleanup_script)
apply_manifest(pp1, catch_failures: true)
end
(which I would like to refactor into a method so all acceptance test files can use it) |
|
I restarted the last successful merge to master build: https://travis-ci.org/github/voxpupuli/puppet-zabbix/jobs/675009991 |
|
whyyyyyyyyyyyyyyyyyyyyyyyyyyy does it pass |
|
I created a simple PR against master that only touches our README.md: #681 |
|
which passed. Now I cherry-picked the cleanup commit into another PR: |
bastelfreak
force-pushed
the
modulesync
branch
3 times, most recently
from
1884f3f
to
64b0344
Compare
|
@baurmatt I'm still debugging this, but if you've any ideas why this works on master but fails here, please let me know :) |
Without this change it's not possible to execute tests in random orders / the agent_spec.rb after any other acceptance test.
Turns out our acceptance tests fail in vagrant because the default centos image has selinux enabled. The docker image on travis not. The is an selinux boolean to allow the httpd service to speak to a database. By default this is of. This PR fixes this.
|
I've no clue why the code in docker fails to create a database for one zabbix version: I'm unable to reproduce this locally with vagrant |
bastelfreak
force-pushed
the
modulesync
branch
4 times, most recently
from
8445187
to
47caa55
Compare
|
@bastelfreak Ha! I've seen something similar with Ubuntu 18.04. Turned out that it ships a This led to /usr/share/doc/zabbix--pgsql-3.4/*.sql not being installed and hence failing tests. I think the problem here is, that |
| c.formatter = :documentation | ||
| hosts.each do |host| | ||
| if host[:platform] =~ %r{el-7-x86_64} && host[:hypervisor] =~ %r{docker} | ||
| on(host, "sed -i '/nodocs/d' /etc/yum.conf") |
There was a problem hiding this comment.
@bastelfreak This shitty thing! :D
There was a problem hiding this comment.
I leads to files in /usr/share/doc/* not being installed. As Zabbix places their database initialization script there, this breaks our acceptance CI.
There was a problem hiding this comment.
Ah yes, I now see the IRC conversation. Pointing @bastelfreak in the right direction.
There was a problem hiding this comment.
Thanks @baurmatt . We figured the same out in parallel yesterday on IRC. The current code should now work again on docker images. Are you already working on Ubuntu 18/20 support for the module?
bastelfreak
changed the title
modulesync 2.12.0