Updates to fluentd from drycc/fluentd

[Cryptophobia]

• chore(fluentd): rename parts from deis to hephy 3861032
• chore(charts): adding external docker_registry support b6885dd
• chore(.travis.yml): change .travis.yml 4199cfc
• chore(fluent): upgrade fluent to v1.4 3898e76
• chore(fluentd): upgrade fluentd 771cb43
• chore(fluentd): add Gemfile.lock 38fb9c8
• fix(influxdb): fix influxdb host and port a17a8b0
• feat(fluentd): support containerd log format e68c723
• fix(charts): skipped value for daemon_environment: Not a table c928f0b
• feat(nsqd): add stateless nsqd cluster support 976d7eb
• chore(influxdb): change influxdb service name 8fbaf60

[kingdonb]

Why does this only fail sometimes?

DigitalOcean CNI "deis logs" works fine in three out of five buildpack and "go/python/http/dockerfile" examples I've tried so far, so it appears the problem doesn't manifest on this cluster. This is on Hephy Workflow v2.22.1

I'll move over to Azure next, so I can see if it repro the issue there. Noted that K8s 1.17.13 is the cluster version that I'm testing on. A lot of our strategies for accessing the host systems via DaemonSet are called out by the cluster linter in the upgrade process, next step testing on 1.18.10.

I saw there was a lot going on in this patch and I just thought it was strange this only apparently affects some clusters, if it's really a (misspelling in the config file class of error) hard misconfiguration then why does it still work sometimes apparently without issue? Who can say for sure...

[kingdonb]

It's definitely not an issue on DigitalOcean. Moving on to test Azure, where I think I was able to reproduce the problem before.

[Cryptophobia]

@kingdonb , I suspected this has to do with using ENV variables inside the container for fluent.conf on the base images that fluentd provides.

I think this FLUENTD_CONF env variable disappeared in one of the newer images from Fluetnd docker hub.