Skip to content

telemetry: cut perf-devnet-3 edge logs over to Vector→OTLP, drop Loki#46

Open
samcm wants to merge 1 commit into
masterfrom
vector-otlp-cutover
Open

telemetry: cut perf-devnet-3 edge logs over to Vector→OTLP, drop Loki#46
samcm wants to merge 1 commit into
masterfrom
vector-otlp-cutover

Conversation

@samcm
Copy link
Copy Markdown
Member

@samcm samcm commented Jun 4, 2026

Cuts perf-devnet-3 edge log shipping off Loki onto Vector's OTLP logs pipeline to the prod analytics gateway (clickhouse-raw external.otel_logs), matching the cutover already live on blob/bal/glamsterdam-devnets.

  • docker_logs → OTLP /v1/logs, log level parsed into OTel severity at ingest
  • ingress identity (auth user + ingress_user tag) = ethereum_network_name (perf-devnet-3), not the sops username
  • Vector → 0.56.0 (keeps the docker.ethquokkaops.io/dh/ mirror prefix)
  • logs-only — no client trace wiring, so no client roll; deploy is --tags otelcol,vector (Vector container restart only)
  • otelcol_contrib_cleanup: true is kept so the deploy removes the currently-running hand-added otelcol on these nodes. Follow-up once confirmed: drop the otelcol_contrib role from the playbook (as done for the other devnets).

Heads-up for whoever runs perf-devnet-3: the committed devnet-3 inventory.ini currently lists only the bootnode — please make sure it reflects the live hosts (the cluster's other nodes appear under the stale devnet-2 inventory) before rolling, and deploy one node at a time. Config-only PR; not deployed.

Replace the Loki log sink (and the hand-added otelcol-contrib experiment)
with Vector's OTLP logs pipeline to the prod analytics gateway, matching
the cutover already shipped on blob/bal/glamsterdam-devnets. Log level is
parsed into OTel severity at ingest.

Ingress identity (auth user + ingress_user tag) is derived from
ethereum_network_name instead of secret_loki.username, so attribution
stays correct regardless of the sops username. Bump Vector to 0.56.0
(keeping the docker.ethquokkaops.io/dh mirror prefix).

otelcol_contrib_cleanup is left true so deploying this removes the
currently-running otelcol container (these nodes have it live). Once a
deploy confirms otelcol is gone, the otelcol_contrib role can be dropped
from the playbook as a follow-up (as done for the other devnets).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant