Description of problem:

It's the same reported in LOG-5466 that should be fixed in Logging 5.9.2, but reviewing the code for Logging 5.9.10, it doesn't contain the fix [0] for when the collector type is fluentd leading to not honour the "no_proxy" configuration available in the cluster-wide proxy.

Version-Release number of selected component (if applicable):

5.9.10
collector type fluentd
output type: HTTP

How reproducible:

Always

Steps to Reproduce:

Logging 5.9.10 doesn't contain the fix [0] for bug LOG-5466 [1]:

Let's install Logging 5.9.10 that should be expected to contain the fix [0] for bug LOG-5466 [1]:

$ oc get csv |grep logging
cluster-logging.v5.9.10           Red Hat OpenShift Logging                        5.9.10      cluster-logging.v5.9.10                      Succeeded

Let's review that the collector type is fluentd:

$ oc get clusterlogging instance -o jsonpath='{.spec.collection.type}' 
fluentd

Let's verify that the fix [0] for taking into consideration the "no_proxy" configuration is not in the file `/usr/share/ruby/net/http.rb`:

$ pod=$(oc get pods -l component=collector -o jsonpath='{.items[0].metadata.name}')
$ oc rsh $pod cat /usr/share/ruby/net/http.rb > /tmp/5910_http.rb

It's visible that the fix [0] is not present in the source file `/usr/share/ruby/net/http.rb` inside the fluentd container:

$ grep "http = new(address, port" /tmp/5910_http.rb 
      http = new(address, port, p_addr, p_port, p_user, p_pass)
      
$ grep "if p_addr && p_no_proxy" /tmp/5910_http.rb 
        if p_addr && p_no_proxy && !URI::Generic.use_proxy?(p_addr, p_addr, p_port, p_no_proxy)

Logging 5.8.16 contains the fix [0] for bug LOG-4784 [4]

$ oc get csv|grep logging
cluster-logging.v5.8.1           Red Hat OpenShift Logging                        5.8.1      cluster-logging.v5.8.0                      Succeeded

$ pod=$(oc get pods -l component=collector -o jsonpath='{.items[0].metadata.name}')
$ oc rsh $pod cat /usr/share/ruby/net/http.rb > /tmp/581_http.rb

Let's see that the fluentd collect pod in Logging 5.8.16 contains the fix [0]

$ grep "http = new(address, port" /tmp/581_http.rb 
      http = new(address, port, p_addr, p_port, p_user, p_pass, (ENV['NO_PROXY']||ENV['no_proxy']))

$ grep "if p_addr && p_no_proxy" /tmp/581_http.rb 
        if p_addr && p_no_proxy && !URI::Generic.use_proxy?(address, address, p_port, p_no_proxy)

Actual results:

Logging 5.9.10 with fluentd as collector type and configured the OpenShift cluster with the cluster-wide proxy doesn't honour the no_proxy options

Expected results:

Logging 5.9.10 with fluentd as collector honours the OpenShift cluster-wide no_proxy options

[0] https://github.com/ViaQ/logging-fluentd/blob/v1.16.x/fluentd/log4366_ruby_lib_net_http.patch
[1] https://issues.redhat.com/browse/LOG-5466
[2] https://github.com/ViaQ/logging-fluentd/blob/3d1274cb223f2f421db47598bff7a35bcbbd9f3e/fluentd/log4366_ruby_lib_net_http.patch#L20
[3] https://github.com/ViaQ/logging-fluentd/blob/3d1274cb223f2f421db47598bff7a35bcbbd9f3e/fluentd/log4366_ruby_lib_net_http.patch#L18
[4] https://issues.redhat.com/browse/LOG-4784