Two posts ago we opened the PagerDuty vertical and found three-of-three alternatives not alive. One post ago we opened Linktree and found two-of-three. We promised a positive-side post before the series got too doomy. Here it is. On /datadog/ we list three self-hostable alternatives — SigNoz, the Grafana LGTM stack, and Uptrace. As of today, three of three are alive. Two of them shipped commits this morning.

That's not a coincidence and it's not because we got lucky on the category. It's the first failure mode in this series where the failure mode is absent — and the thing that holds it back, the structural reason this vertical doesn't decay, is worth the post.

The three we list (and what their health pill says today)

Repo	Stars	Last commit	Archived	State
SigNoz/signoz	26,857	2026-05-10	no	alive
grafana/grafana	73,650	2026-05-10	no	alive
uptrace/uptrace	4,196	2026-04-30	no	alive

Same exact freshness query we ran on PagerDuty and Linktree. Different answer, because the surrounding economics are different.

SigNoz — the commercial-OSS playbook executed correctly

SigNoz/signoz shipped a commit at 09:27 UTC today. The repo has a managed-cloud tier (paid), an Enterprise edition (paid), and a fully-functional MIT community edition (free). Bessemer-backed Series A in 2022, headcount visibly growing on commit history, ClickHouse-based architecture that maps onto whatever retention budget you give it.

This is the model on-call OSS couldn't make work and link-in-bio OSS couldn't fund — same idea, different category gravity. SigNoz monetizes hosting + Enterprise features for the customers that need them, while the self-host story stays a first-class product because the self-host customer is a credible upgrade path to the paid tier. PagerDuty's "ecosystem" did not contain a single vendor willing to run that play. Observability has SigNoz, Grafana Labs, Tempo's predecessor (Jaeger), and the broader OpenTelemetry community all aligned on it.

Grafana stack (Mimir + Loki + Tempo) — the platonic ideal of OSS that scaled

grafana/grafana shipped a commit at 10:40 UTC today. This is a company-grade open source codebase: 73,650 stars, hundreds of contributors, AGPL core, three sibling projects (Mimir for metrics, Loki for logs, Tempo for traces) each with their own tens of thousands of stars and active release cadence.

Grafana Labs is the existence proof that OSS observability can fund itself at scale. They have a paid hosted product (Grafana Cloud), an Enterprise on-prem product, and a community edition that runs the same code minus a small set of commercial features. Grafana OnCall — the project we wrote about being archived last week — sits inside the same company, which is the wrinkle. Grafana the company is alive and well; Grafana the on-call product was a strategic archive-and-rebrand into Grafana IRM (paid). The observability core is not under the same pressure: there is no managed-Datadog acquisition target large enough to justify retiring Mimir, Loki, or Tempo behind a paywall.

Uptrace — the smaller-footprint alternative still finds room to breathe

uptrace/uptrace last shipped April 30. AGPL-3.0, Go server, ClickHouse + Postgres backend. 4,196 stars — small relative to SigNoz, but the repo is healthy: regular tagged releases, single-maintainer-plus-contributors cadence, no archive flag, no signs of acquihire pressure.

Uptrace exists because the SigNoz shape (OpenTelemetry + ClickHouse) is now a category of project, not a single product. You can pick the one whose ops footprint matches your team, swap your collector config, and you're done. The category supports more than one entrant because the data shape is open and the switching cost between them is hours, not weeks.

Why observability didn't decay like on-call did

Three structural reasons, in roughly the order they matter.

1. OpenTelemetry won the data-shape war. Five years ago, every observability vendor had a proprietary agent and proprietary wire format — Datadog's dd-trace, New Relic's, Dynatrace's, Splunk's, all mutually incompatible. Today, OpenTelemetry is the de facto standard for traces, well on its way for metrics, with logs catching up. CNCF graduation, Microsoft / AWS / Google all shipping OTel as the preferred path. Datadog itself documents the OTel ingest path because their customers demanded it. That's the wedge: when the data shape is shared, switching off the SaaS is an agent-config change, not a re-instrumentation project. Every self-host Datadog alternative on our /datadog/ page speaks OTel natively. The vendor lock-in that used to make these migrations cost months of engineering time is mostly gone.
2. The pricing pain is too universal to ignore. Coinbase's $65M Datadog bill made headlines in 2022; it became a meme but it described a real pattern. Per-host pricing in the $15-23/host/month range, plus per-GB log ingest, plus per-event custom-metric overage, compounds nonlinearly with how successful your business is. There's a constant supply of teams who hit a usage cliff and start a Datadog escape project. That demand keeps SigNoz / Grafana / Uptrace funded — both directly (paid tiers) and indirectly (visibility, usage, contributors).
3. Commercial-OSS works in observability because operators value the same thing customers do. The buyer of a paid tier of SigNoz is a senior SRE. The maintainer of SigNoz is a senior SRE. They want the same product. Compare link-in-bio: the buyer of a Linktree paid tier is a creator who doesn't want to think about HTML; the maintainer of LittleLink is a developer who has already written the HTML. Mismatched audience, no funding feedback loop. Observability's buyer-maintainer alignment is the structural feature.

What's actually missing from our /datadog/ page

Editorial debt, in the format we've used for the prior two vertical posts:

• SigNoz's weak_at note understates the project. Our page currently says "RUM, synthetic monitoring, and the Datadog-style alerting DSL are missing or thinner." All three were genuinely missing or thin in 2024. As of mid-2026 SigNoz ships logs (parity), exceptions tracking (parity), and a recently-added alerts query builder. RUM is still on the roadmap rather than shipped, but the gap is narrower than our note implies. The next sweep should rewrite the weak_at field and probably move SigNoz from "good fit for OTel teams" to the lead recommendation for most teams.
• We don't list jaegertracing/jaeger. Jaeger is the original distributed-tracing project, CNCF-graduated, still alive (jaegertracing/jaeger-v2 ships actively), and a perfectly good drop-in if all you want is traces. We left it off because it's traces-only and the page is framed around full Datadog parity. That's defensible but worth adding as a "trace-only" alternative card. June sweep candidate.
• Cost framing is the most-asked question and we answer it weakly. The migration guide hand-waves "ClickHouse needs 4GB+ RAM minimum" and stops. A team currently paying $4,000/month to Datadog wants to know whether their self-host bill is $50, $500, or $5,000. The answer depends on retention and host count and we should tabulate it the way we tabulate setup-time scores. Adding a per-tier cost table is queued.

We hold off on rewriting /datadog/ right now for the same reason we held off on /pagerduty/ and /linktree/: the page being a generation behind is the post's wedge. The June sweep ships the corrections together with the month-over-month diff post.

Where this leaves the series

Three vertical posts in, three distinct failure-mode shapes:

• On-call — OSS dying under vendor pressure. PagerDuty alts at 3-of-3 not-alive; the largest player (Grafana OnCall) was archived in favor of a paid Grafana IRM rebrand. Successor (keephq/keep) added.
• Link-in-bio — OSS dying under vendor indifference. Linktree alts at 2-of-3 not-alive; no commercial OSS plausible because the buyer-maintainer mismatch breaks the funding loop.
• Observability — OSS thriving on a shared data standard. Datadog alts at 3-of-3 alive; OpenTelemetry made the migration mechanical and the buyer-maintainer alignment makes commercial OSS work.

The combined picture is more useful than any single post. A self-host directory that ranks by stars promotes the wrong projects, because stars accumulate slowly and don't decay when a project dies — but freshness alone isn't the whole picture either. The structural question is whether the category has a working commercial-OSS feedback loop. Where it does (observability), self-host alts compound. Where it doesn't (on-call, link-in-bio), they decay at different speeds. That framing is what os-alt is increasingly about, and it's what the graveyard page and the public JSON API are designed to surface programmatically: every alternative we recommend, with a freshness pill that reflects the live commercial reality of the maintainer behind it, not just the GitHub stars.

If you trusted our /datadog/ page last week

The page is in better shape than /pagerduty/ or /linktree/ were when we wrote about them — every alt is alive, every freshness pill is green, the migration guide is concrete. The directory bug is mostly that we underrate SigNoz's completeness; the recommendation table is editorially correct. If you were planning a Datadog escape and our page is the starting point, the short answer is SigNoz for most teams (lighter ops, single binary architecture, OTel-native), the Grafana LGTM stack if you already have an SRE who enjoys it, and Uptrace if you want a smaller-footprint variant of the SigNoz shape. All three will still be shipping next month.

What's next

This is the third vertical-deep post in the series. Coming up:

• The first month-over-month diff (June 2026 — what flipped state, including whether LinkStack went stale on schedule)
• Form builders post-Typeform (ohmyform abandoned — what's actually shipping for self-host forms)
• A license-drama vertical (where the OSS is technically alive but the license changed — Sentry's BSL, MongoDB's SSPL, ElasticSearch / OpenSearch)

If you'd rather get the diff post in your inbox than rediscover us via search later, the form below ships the monthly digest the same day each post lands.