> Source: https://glassmkr.com/for-sre # Monitoring for SREs who manage bare-metal fleets. Opinionated alert rules. Open-source agent. Pricing you can predict. Built by infrastructure operators, for infrastructure operators. [Install in 2 minutes](https://app.glassmkr.com/register) [View Crucible on GitHubMIT](https://github.com/glassmkr/crucible) 3 nodes free. No card required. `curl -fsSL https://app.glassmkr.com/install.sh | bash` Crucible v0.13.20 on npm ## The problem If you run bare-metal infrastructure, you already know what should be monitored. The frustrating part isn’t figuring out what to watch. It’s the operational overhead of getting a monitoring tool set up that actually catches the things that matter without burying you in noise. The big platforms (Datadog, New Relic) are designed for ephemeral cloud workloads. Their default rules don’t know what a degraded RAID looks like, what ECC errors mean at scale, or when an NVMe drive is about to fail. Their pricing scales with telemetry volume, which is exactly the wrong shape for bare metal. The open-source stack (Prometheus, Grafana, Alertmanager) works, but it’s a three-week project to deploy and a permanent maintenance burden. You’re not running a monitoring engineering team. You’re running infrastructure. ## How Glassmkr fits Glassmkr is monitoring built around the assumption that you’re running bare-metal hardware and you know what you’re doing. ### Opinionated defaults that match real failure modes. 65 alert rules tuned for bare metal: disk SMART, RAID state, NVMe wear, IPMI sensors, ECC errors, ZFS health, network interface state, conntrack exhaustion, BMC accessibility, kernel update pending, security patches available. The thresholds are tuned for “this is genuinely a problem worth waking someone up about,” not “CPU briefly spiked.” ### A 150-line bash installer. Read the install script before you run it. Read the agent source on GitHub before you install it. The agent runs as a non-root user, communicates over HTTPS only, and ships exactly the metrics it claims to ship. ### Multi-channel notifications routed how your team works. Telegram, Slack, Discord, PagerDuty, email, generic webhook. Group by team, by server, by severity. Suppress during maintenance windows. The alerting layer doesn’t impose a workflow on you. ### Remediation guidance on every alert. Every alert detail page shows the recommended next step: what to verify, what command to run, what to expect after. Cross-references between related rules (e.g. NVMe wear suggests pre-failure capacity planning before it becomes a SMART critical). ### Predictable pricing. $3 per node per month after 3 free nodes. No telemetry volume billing. No tier upgrades for the features your team actually uses. Every Pro account gets every feature. ## The signals on-call reads first A firing alert with the exact remediation, plus per-interface saturation against the link ceiling, hardware errors and discards. The page leads with what is wrong and what to run, not a wall of graphs. app.glassmkr.com/server/edge-sjc-02 Alerts(1) warningNIC errors on eno21h 4m eno2 has accumulated 160 hardware errors (CRC + frame) and 90 discards. Error ratio exceeds 0.1% of packets. Check cabling / SFP. `interface_errors` Networksaturation · errors · discards eno11 GbE Saturation19.2% Errors0 Discards0 eno21 GbE Saturation4.8% Errors160 Discards90 History1h6h24h7d30d Memory used3.7 / 16.0 GB ## Specific alert rules that matter The 65 rules most relevant to bare-metal SRE work are roughly: - **Storage health**: disk SMART degradation, NVMe wear bands, disk I/O errors, disk latency anomalies - **RAID state**: degraded arrays, scrub errors, rebuilds in progress - **ZFS health**: pool state, vdev errors, scrub recency - **Filesystem state**: read-only mounts, inode exhaustion, file descriptor exhaustion, capacity thresholds - **Memory pressure**: RAM utilization, swap utilization, OOM kills observed - **CPU state**: sustained load, sustained I/O wait, temperature anomalies - **Network state**: interface errors, link speed mismatches, saturation, bond state, conntrack exhaustion - **Hardware sensors (IPMI)**: fan failures, PSU redundancy loss, SEL critical entries, BMC accessibility - **System health**: clock drift, NTP sync state, systemd service failures, unexpected reboots - **Security state**: pending security updates, kernel vulnerabilities flagged, SSH password authentication exposed, no firewall configured Full list at [/docs](/docs). Every rule has a tunable threshold, dashboard-rendered remediation guidance, and clear evidence about why it fired. ## The agent you can read Crucible is MIT licensed and on npm. The source is at [github.com/glassmkr/crucible](https://github.com/glassmkr/crucible). The install script is ~150 lines of bash. The agent runs as the `glassmkr` user, never root. It writes to its own systemd unit. It communicates with the dashboard over HTTPS only. It ships metrics and alert state; it does not ship logs, command output, or anything secret. If you can’t read the agent, you can’t trust it. We made the agent readable. ## Pricing reminder $3 per node per month. First 3 nodes free. No card required to install. Cancel any time. - For 10 nodes: **$21/month** - For 50 nodes: **$141/month** - For 200 nodes: **$591/month** [Install now](https://app.glassmkr.com/register) [See full pricing page](/pricing) ## If it fits, install on a single server first. If you want to see if it works on your stack, install on a single server in 2 minutes: `curl -fsSL https://app.glassmkr.com/install.sh | bash` [Sign up free](https://app.glassmkr.com/register) Questions about whether Glassmkr fits your environment? Email [simon@glassmkr.com](mailto:simon@glassmkr.com) directly.