It feels a bit clunky but it's finally working.
I did not find any other simpler combination than this.

Refs. swh/infra/sysadm-environment#4955

It's unfortunately not working the same as main clusters.
The data is sent json like already...

Refs. swh/infra/sysadm-environment#4955

Refs. swh/infra/sysadm-environment#4955

To avoid parsing log issues (from services which generate system logs).

Refs. swh/infra/sysadm-environment#4955

Refs. swh/infra/sysadm-environment#4955

As on some machines, this prevents from running.

Refs. swh/infra/sysadm-environment#4955

To match what is done in the remaining part of the readme.

Refs. swh/infra/sysadm-environment#4955

Through the cli.

Refs. swh/infra/sysadm-environment#4955

Through the cli.

This is necessary for the log parsing filtering to happen properly (it uses
the namespace to filter data). The namespace used so far, default, was filled
with other pods (whose logs could not be parsed the same way, creating
divergence in what's currently deployed in "production" clusters).

Refs. swh/infra/sysadm-environment#4955

It's used within the Chart.yaml as conditional and it was not installed
properly without this (resulting in issues when trying to install the elk
part).

Refs. swh/infra/sysadm-environment#4955

Refs. swh/infra/sysadm-environment#4955

It seems doing the manual install of the crds is not really needed

Antoine R. Dumont authored 1 year ago and Vincent Sellier committed 1 year ago

It's too bad the helm dependency build does not reuse the Charts.lock which
contains all this already, a bootstrap issue I presume.

Install all the necessary components to run a swh environment
it's a WIP / POC as it's not completely clear if it will converge
with the cluster-configuration or need a sort of orchestrator.

This first commit allow to start an swh-storage

and disable the postgresql first instance

Related to swh/infra/sysadm-environment#4879

Refs. swh/infra/sysadm-environment#4971

The replayer's code is not calling storage.flush() so a part
of the replayed content is lost when the replayer restart
or is taking a loooonnggg time to be replayed, specially low
occurences objects like snapshot with a default buffer size of 25 000.

Related to swh/infra/sysadm-environment#4879

to allow to debug

Related to swh/infra/sysadm-environment#4824

Start to test a real workload with cassandra as main backend

Related to swh/infra/sysadm-environment#4824