---
title: Time gaps with missing data
source: https://docs.newrelic.com/docs/infrastructure/infrastructure-troubleshooting/troubleshoot-infrastructure/time-gaps-missing-data
---

## Problem

Your New Relic infrastructure agent is up and running, but when your CPU reaches a high percentage of usage, the agent stops intermittently to submit data.

## Solution

**Required agent version**: 1.0.1002 or later

The following steps mitigate the data gaps problem:

1.  Remove the limit of one thread for the agent, allowing one thread per core.

    **Edit the newrelic-infra.yml configuration file (Windows)**

    1.  Open the following file in your favorite plain text editor: `C:\Program Files\New Relic\newrelic-infra\newrelic-infra.yml`
    2.  Add the following configuration option:

        ```yml
        max_procs: -1
        ```

    **Edit the newrelic-infra.yml configuration file (Linux)**

    1.  Open the following file in your favorite plain text editor: `/etc/newrelic-infra.yml`
    2.  Add the following configuration option:

        ```yml
        max_procs: -1
        ```
2.  Use [your init system](https://docs.newrelic.com/docs/infrastructure/new-relic-infrastructure/configuration/start-stop-restart-check-infrastructure-agent-status#init-system) to restart the agent service:

    **Restart the agent with SystemD**

    Use SystemD commands with CentOS 7, Debian 8, RHEL 7, and Ubuntu 15.04 or higher:

    ````sh
    sudo systemctl restart newrelic-infra
    ```

    ````

    **Restart the agent with System V**

    Use System V commands with Debian 7:

    ````sh
    sudo /etc/init.d/newrelic-infra restart
    ```

    ````

    **Restart the agent with Upstart**

    Use Upstart commands with Amazon Linux, CentOS 6, RHEL 6, and Ubuntu 14.10 or lower:

    ````sh
    sudo initctl restart newrelic-infra
    ```

    ````

    **Restart the agent in Windows**

    ````sh
    net stop newrelic-infra
    net start newrelic-infra
    ```

    ````

## Cause

The New Relic infrastructure agent, by default, runs in a single operating system thread. This may mean that, especially in Windows environments, the process scheduler gives it little chance to get CPU time when the system is overloaded.

The `max_procs: -1` configuration option removes this limitation and allows the agent to use one thread per CPU core (as maximum).