aleks_me2

I have decided to use clickhouse with that config because of missing S3 for logs and metrics for long term store.

  <clickhouse>
    <storage_configuration>
      <disks>
        <audit_s3>
          <type>s3</type>
          <endpoint>https://S3-EndPoint/{{ audit_bucket_name }}/clickhouse/</endpoint>
          <access_key_id>{{ clickhouse_audit_s3_access_key }}</access_key_id>
          <secret_access_key>{{ clickhouse_audit_s3_secret_key }}</secret_access_key>
        </audit_s3>
      </disks>
      <policies>
        <audit_tiered>
          <volumes>
            <default>
              <disk>default</disk>
            </default>
            <audit_s3>
              <disk>audit_s3</disk>
            </audit_s3>
          </volumes>
        </audit_tiered>
      </policies>
    </storage_configuration>
  </clickhouse>

To your point about it maybe being a paid feature: the DB-IP Lite files are free (CC-BY), so the ASN part doesn't have to cost anything — it's just not what the MaxMind-oriented tooling usually reaches for.

For ad hoc analytics these tools are very nice, but for long-term stats they run into timeouts, because the backend needs a lot of time to collect the data and build the reports for ~1 month or a quarter — or am I wrong about that?

What I'm seeing matches the observations of the other commenters here exactly: most of the "users" are bots harvesting pages and PDFs. My solution for getting at the real users is GoatCounter.

I use it to create pdfs for my blog posts with that code :-)

  python3 "$SCRIPT_DIR/preprocess.py" > blog-post.md && \
  /datadisk/Downloads/pandoc-3.9/bin/pandoc \
    blog-post.md \
    --pdf-engine=typst \
    --template="templates/blog.typ" \
    -o "out.pdf" 2>/tmp/pdf-pandoc-err.txt;