Written by: Nimesh Chakravarthi, Co-founder & CTO, Struct
Key Takeaways
- Free on-call stacks combine open-source collectors, notification layers, and incident tools to deliver zero-cost paging for small teams.
- Grafana OnCall OSS was archived in March 2026, which pushes users toward paid Grafana Cloud IRM or self-maintained forks.
- OneUptime and GoAlert provide the most self-contained free options for escalation and scheduling, while SigNoz and Zabbix add deeper observability at higher setup cost.
- Hidden costs such as infrastructure, SMS fees, and engineering maintenance time often outweigh the initial savings of free tools as alert volume grows.
- Struct automates root-cause investigation and runbook execution so teams can move beyond manual triage, see how Struct works.
1. Grafana OnCall OSS After the 2026 Archive
Grafana OnCall’s OSS version entered read-only maintenance mode in March 2025 and was archived on March 24, 2026, with Grafana merging OnCall and Incident into a unified Grafana Cloud IRM app. Self-hosted deployments still run, but receive no new features or security patches. Teams already on the OSS version must either migrate to Grafana Cloud IRM or fork the repository. The Docker image remains publicly available, but production use carries increasing operational risk as the codebase ages without upstream support.
For teams evaluating the archived version in a test environment, the Docker image remains publicly available. Quick-start reference (archived image, use at your own risk):
docker run -d --name grafana-oncall -p 8080:8080 grafana/oncall:latest
Hidden cost warning: Grafana Cloud IRM’s free tier caps alert rules and active users, so teams exceeding those limits face immediate upgrade pressure.
2. SigNoz Free APM and Webhook Limits
SigNoz is an open-source APM and observability platform built on OpenTelemetry. The self-hosted Community Edition covers distributed tracing, metrics, and log management with no seat fees.
SigNoz Pricing and Feature Limits
The self-hosted Community Edition is free indefinitely. However, webhook-based alert channels are limited in the community build, and advanced features such as SSO, SAML, and extended data retention sit behind the paid SigNoz Cloud or Enterprise tiers. Teams running high alert volumes also absorb the infrastructure cost of the ClickHouse backend, which scales non-trivially. Self-hosting SigNoz requires managing ClickHouse, the query service, and the frontend as separate containers, which places it in the high integration-complexity bracket alongside the full Grafana stack.
docker compose -f docker/clickhouse-setup/docker-compose.yaml up -d
3. OneUptime All-in-One Monitoring and On-Call
OneUptime is a fully open-source, self-hosted platform covering uptime monitoring, on-call scheduling, status pages, and incident management in a single repository. It is one of the few free tools that ships escalation policies, on-call rotations, and a status page out of the box.
OneUptime Compared to Grafana OnCall
Grafana OnCall, when it was active OSS, required Grafana, Prometheus, and Alertmanager as prerequisites before OnCall added paging on top. OneUptime ships the entire stack in one Docker Compose file, which reduces initial setup friction significantly. The tradeoff is that OneUptime’s APM and log analysis capabilities are shallower than Grafana’s mature observability ecosystem. For teams whose primary need is escalation and status pages rather than deep telemetry, OneUptime is the more self-contained choice.
git clone https://github.com/OneUptime/oneuptime && cd oneuptime && cp config.env.example config.env && docker compose up -d
4. GoAlert for Pure Paging Needs
GoAlert Open Source Capabilities
GoAlert is an open-source on-call scheduling and alerting tool developed by Target. It focuses exclusively on paging, including on-call schedules, escalation policies, and multi-channel notifications (SMS, voice, Slack, webhook). It does not collect metrics or ingest logs, so it must be paired with a separate monitoring tool such as Prometheus Alertmanager or Zabbix. GoAlert fits teams that already have observability covered and need only a reliable, self-hosted paging layer.
docker run -d -p 8081:8081 -e GOALERT_DB_URL=postgres://user:pass@host/goalert goalert/goalert
Hidden cost warning: SMS and voice notifications route through Twilio or similar providers, which adds per-message costs that accumulate quickly during high-alert periods.
5. Zabbix for On-Call and Infrastructure Monitoring
Zabbix is a mature, fully open-source monitoring platform with built-in alerting, escalation scripts, and media types (email, Slack, PagerDuty webhook). It covers infrastructure, network, and application monitoring without a license fee. The on-call workflow relies on action-based escalation rules configured in the Zabbix UI, which are powerful but require significant YAML and UI configuration time. Open-source SRE tools like Zabbix are free of license cost but require significant operational expertise for scaling, configuration, and maintenance, which creates long-term sustainability tradeoffs beyond initial setup.
docker run --name zabbix-server -e DB_SERVER_HOST=mysql-server -e MYSQL_USER=zabbix -e MYSQL_PASSWORD=zabbix_pwd -d zabbix/zabbix-server-mysql:latest
6. Better Stack Free Tier for Lightweight Monitoring
Better Stack’s free plan provides 10 monitors with 3-minute checks plus a status page, targeting small teams with simple infrastructure monitoring needs. On-call scheduling and escalation policies are available on paid tiers. The free tier works as a starting point for a single-service startup but hits hard limits quickly as the number of monitored endpoints grows. Better Stack is a managed SaaS product, so there is no self-hosting option, and all data leaves your infrastructure.
On-call rotations, SMS alerts, and phone calls require a paid plan. Teams treating the free tier as a long-term solution encounter a forced upgrade at the worst possible moment, during an incident.
7. 2026 Project Status Updates
Several projects in this space changed status between 2025 and 2026. Grafana OnCall OSS was archived March 24, 2026, and Grafana Cloud IRM replaced it as the supported path. SigNoz Community Edition continues as an actively maintained project. OneUptime, GoAlert, and Zabbix also remain active open-source projects with recent commits. Better Stack rebranded from Logtail/Uptime.com and consolidated its product line in 2025. Given this level of churn across the ecosystem, teams evaluating any open-source tool should verify the last commit date and open issue velocity on GitHub before committing to a production deployment.
The following table summarizes the current state of each tool’s core capabilities and the hidden costs teams should anticipate before deployment.
Free On-Call Capabilities Comparison Table
| Tool | Alerting + Escalation | Self-Hosting Difficulty | Hidden Cost Warning |
|---|---|---|---|
| Grafana OnCall OSS | Yes (archived, no patches post-Mar 2026) | High, requires full Grafana stack | Cloud IRM paid tiers for active support |
| SigNoz Community | Yes (webhook limits on free) | High, ClickHouse and multi-container | SSO, retention, advanced alerts are paid |
| OneUptime | Yes, escalation and status page included | Medium, single Docker Compose | Shallow APM, compute costs scale with data |
| GoAlert | Paging only, no metrics collection | Low, single binary or Docker | Twilio SMS and voice costs per message |
| Zabbix | Yes, action-based escalation | Medium to high, UI and DB configuration | Engineering time for ongoing maintenance |
| Better Stack (free) | 10 monitors, no on-call rotation | N/A, managed SaaS only | On-call and SMS require paid plan |
Each of these tools presents deployment and scaling challenges that do not appear in a simple feature grid. The next section highlights the setup issues that teams most often encounter in real-world use.
Troubleshooting Common Setup Issues
The following pain points appear repeatedly in community threads and Reddit discussions about self-hosted on-call stacks.
- Grafana OnCall webhook loops: After the March 2026 archive, teams report that the OSS alertmanager-to-oncall webhook integration silently drops alerts when the OnCall worker container restarts. The fix is to pin the container restart policy and add a health-check endpoint probe.
- SigNoz ClickHouse disk exhaustion: Default retention settings fill disk within days on high-traffic services. Set
data_retention_in_daysexplicitly in the ClickHouse config before first run. - GoAlert SMS delivery failures: Twilio trial accounts block international numbers and throttle burst sends. Production deployments need a verified Twilio sender ID and a paid account before go-live.
- Zabbix action escalation not firing: The most common cause is a misconfigured “Default time period” in the escalation step. Zabbix requires at least one step with a non-zero delay before the next step triggers.
- OneUptime probe container connectivity: Probes deployed behind a NAT without outbound port 443 whitelisted silently fail health checks. Verify egress rules before assuming the monitor is working.
When to Upgrade from Free Tools
The 2025 SRE Report from Catchpoint noted that toil increased in 2025 after prior declines. Free tools handle routing and paging well. They do not handle investigation.
The signal that a free stack has become a liability rather than an asset typically falls into two categories. First, volume: when a senior engineer’s week is consumed entirely by reactive triage rather than product work, the cost of their time exceeds any savings from avoiding a commercial tool. Second, severity: when SLA windows are measured in minutes and manual log-hunting across Datadog, CloudWatch, and Sentry consumes 30 to 45 minutes per incident, every alert becomes an SLA risk.
Integration with your existing monitoring and observability toolchain is the single most important functional criterion for incident management platforms, and free tools rarely provide the cross-stack correlation that turns raw alert data into a root cause. The result is tribal knowledge bottlenecks: junior engineers cannot safely take on-call shifts because they lack the systemic context that senior engineers carry in their heads.
Struct: AI-Powered Automated Investigation
Struct is an AI agent that automatically root-causes engineering alerts by pulling and analyzing metrics, logs, traces, monitors, and code. It performs regression analysis, correlates anomalies, and generates impact summaries before an engineer opens their laptop.
Free tools stop at paging, and Struct starts investigating. When an alert fires in a designated Slack channel, Struct immediately queries connected observability sources such as Datadog, AWS CloudWatch, GCP Logs, Sentry, Grafana, and Prometheus. It then correlates the data into a unified timeline and posts a root-cause summary with suggested fixes directly in the thread. Companies like FERMAT and Arcana use Struct to investigate thousands of alerts monthly, with large-scale customers reporting an 80% reduction in triage time.
Setup takes under 10 minutes. You authenticate Slack, connect GitHub, link your observability platform, and auto-investigations begin immediately. Struct is fully SOC 2 and HIPAA compliant, which makes it viable for fintech and healthtech teams with strict data requirements. Custom runbooks can be encoded directly into Struct so the AI investigates exactly as a senior engineer would. Junior engineers then gain a reliable starting point for every alert without an automatic escalation.
Book a demo to automate your on-call runbook
Frequently Asked Questions
Is SigNoz completely free for production use?
SigNoz’s self-hosted Community Edition carries no license fee and covers distributed tracing, metrics, and log management. However, advanced capabilities including SSO, SAML authentication, extended data retention, and higher webhook alert volumes require the paid SigNoz Cloud or Enterprise plan. Infrastructure costs for the ClickHouse backend also scale with data volume, so “free” in practice means absorbing compute and storage costs that grow with traffic. For small teams with modest alert volumes and existing Kubernetes infrastructure, the Community Edition is genuinely production-viable. For teams scaling past a few hundred gigabytes of telemetry per day, the operational overhead and infrastructure bill often exceed the cost of a managed alternative.
What is the best free Datadog alternative for on-call monitoring in 2026?
The answer depends on what part of Datadog’s functionality the team needs to replace. For pure paging and escalation, GoAlert is the most operationally simple self-hosted option. For an all-in-one stack covering uptime, escalation, and status pages, OneUptime ships the most complete feature set in a single Docker Compose deployment. For teams that need APM and tracing alongside alerting, SigNoz Community Edition is the strongest open-source option, though it carries higher setup complexity. None of these tools replicate Datadog’s cross-signal correlation or automated investigation, and that gap is where AI-native tools like Struct operate.
How does GoAlert compare to Grafana OnCall for open-source on-call scheduling?
GoAlert focuses exclusively on on-call scheduling, escalation policies, and multi-channel notifications. It is lightweight, easy to deploy, and does not require a surrounding observability stack. Grafana OnCall OSS, before its March 2026 archival, provided deeper integration with Grafana’s alerting ecosystem and context-rich notifications that included metrics data alongside the page. For teams already running Grafana and Prometheus, OnCall was the more integrated choice. For teams that want only a paging layer without observability dependencies, GoAlert remains the cleaner option. Neither tool performs automated incident investigation, so engineers still need to manually hunt root causes after receiving the page.
What are the hidden costs of running a self-hosted on-call stack?
The most significant hidden cost is engineering time. Configuring escalation rules, maintaining container uptime, upgrading dependencies, and debugging silent failures in webhook integrations consume hours that do not appear on a software invoice. SMS and voice notifications through providers like Twilio add per-message costs that spike during incident storms. Storage and compute for tools like SigNoz’s ClickHouse backend scale with data volume. When a self-hosted tool fails during an incident, the team is simultaneously managing an outage and a broken alerting system. These operational costs explain why many teams eventually migrate to managed or AI-native solutions once their alert volume and SLA pressure cross a threshold.
When should a startup stop using free on-call tools and invest in an automated solution?
The clearest signal appears when senior engineer time spent on triage exceeds the cost of a paid tool. A senior engineer billing at market rate who spends even four hours per week on manual log-hunting represents a significant recurring cost, which compounds as the team grows and alert volume increases. A second signal is SLA pressure: when incident windows are measured in minutes and manual investigation consistently consumes 30 or more minutes per alert, free tools create direct business risk. A third signal is onboarding friction: if junior engineers cannot safely take on-call shifts without escalating to a senior, the team has a tribal knowledge problem that no amount of runbook documentation fully solves. Struct’s automated first-pass investigation addresses all three signals simultaneously, with a 10-minute setup and a 30-day risk-free pilot. As noted earlier, customers report an 80% reduction in triage time once Struct is in place.
