CPU usage and concurrent connections

All our hosting services have limits on CPU, memory, input/output and processes. They exist to isolate each account, keep the service stable and stop one application's usage from affecting everyone else.

In cPanel, look for Resource Usage or CPU and Concurrent Connection Usage. You will find a history showing whether your account was ever limited by these limits, so that you can work on its usage by optimising the applications or moving to a plan with more resources.

Check the period shown in the initial summary and open the details. Adjust the interval and check the time zone before linking a spike to an incident.

The Details and Snapshots options let you dig deeper.

Details

It gives you usage charts for different intervals, so you can get an overall view of resource usage. The most common metrics are:

• SPEED or CPU: percentage of CPU used by the account.
• PMEM: physical memory used by its processes.
• IO and IOPS: throughput and number of disk read or write operations.
• EP (Entry Processes): dynamic requests entering the account at the same time. It is not the same as the number of connected visitors.
• NPROC: processes running at the same time, including processes that do not come from a web visit.

A single spike is not enough to identify the cause. Check which resource hit its limit, when it happened and whether it repeats before changing the application or the plan.

Snapshots

This is the feature to turn to when debugging problems. It provides "captures" of processes, requests and, when available, SQL queries from moments when a limit was reached. These captures help narrow down the source, but you should check them against the logs and the application before blaming a specific plugin, query or visitor for the usage.

Interpret the limit and decide what to do next

Tell apart the observed usage, the maximum allowed and the counters of limits reached (faults) when they are available. Hitting the CPU or input/output limit can slow operations down; running out of memory or processes can cause errors. A single percentage does not identify the plugin responsible.

Note the time, the URL or operation and the resource affected. Compare them with scheduled tasks, imports, backups, visits and logs. If a PHP request is slow, analyse a sample with PHP X-Ray.

If the usage comes from a one-off task, reduce its concurrency or schedule it outside peak activity. If it is sustained and comes from useful work that is already optimised, consider adding resources. Do not buy more capacity just because of an unexplained spike.

When you ask for help, include the interval, the time zone, the limited resource and a capture without sensitive data. A snapshot is a sample, not a complete record of everything that happened.