1 min read
Xpulse monitors the server’s load average and raises an alert when the box is oversubscribed. Two severity levels, one cause: the server is busier than its CPU cores can comfortably handle. The alert reads the 15-minute load average (the smoothest measure) and compares it to the server’s CPU core count, so it is meaningful on any size of box.
| Alert | Severity | Threshold |
|---|---|---|
| Server Load High | ⚠️ warning | 15-min load above 2× the CPU cores |
| Server Load Critical | 🚨 critical | 15-min load above 4× the CPU cores |
Both link to this article.
Load averages come in three windows (1, 5, and 15 minutes). The 1- and 5-minute values spike constantly: a single heavy request, a backup job, or a burst of traffic can push them up for seconds. The 15-minute average smooths that noise and shows whether the server is sustained busy, which is the only state worth alerting on. If the 15-minute average is high, the load is real and has been building.
The alert message says whole-server load deliberately. On a shared server, load average reflects every site and every tenant on the box, not just the site you are looking at. A high load alert does not mean your site caused it: it means the box is busy, and the cause may be another tenant, a background process, or the host’s other workloads.
This matters for two reasons:
Nothing, at first. This is a server-level signal we investigate on our side. The alert tells our team a box needs attention; we check running processes, other tenants, and resource allocation before deciding on a fix.
If you were doing something heavy when the alert appeared (import, migration, media job, backup restore), tell us: it speeds up the diagnosis considerably.
| Level | What it means | Response |
|---|---|---|
| High (2×+ cores) | Server is oversubscribed; processes may slow down | Investigate heavy processes and tenants; monitor |
| Critical (4×+ cores) | Heavily oversubscribed; processes are queueing | Active investigation: find and resolve the load source |
Xponent-managed sites run on isolated, monitored infrastructure. Load alerts are part of our proactive monitoring: they surface a struggling box before sites on it degrade, not after. On shared infrastructure, the fix is usually a resource or tenant action on our side: you rarely need to do anything.
