---
title: "Monitor an instance"
description: "Watch a TIR Instance's CPU, memory, GPU, and storage use, read and download its logs, see its events, and attach alerts."
content_type: "how-to"
product: "tir-instance"
platform: "TIR"
url: "/tir/tir-instance/how-to/monitor-an-instance/"
topic: ["monitor"]
---

# Monitor an instance

Use an instance's monitoring, logs, events, and alerts to catch a full disk before it stops your
work, and to find out why an instance restarted.

## Before you begin

- An instance in the **Running** state. **Monitoring** shows data only while the instance is
  running, and **Logs** and **Instances Event** stay empty while it is stopped or stopping.
  You can attach alerts only while it is running.
- An alert in the project, if you want to attach one. You can create it from the **Alert** tab.

## Watch resource use on the Monitoring tab

Open the instance's page and select **Monitoring**.

Two readings at the top tell you how full the instance's storage is:

- **Workspace Usage** is the workspace disk at `/home/jovyan`. When it nears 100%, increase it on
  the **Workspace Size** tab while the instance keeps running. See
  [Increase the workspace size](/tir/tir-instance/how-to/increase-the-workspace-size/).
- **Ephemeral Storage** is the rest of the container: every folder outside `/home/jovyan` and your
  mounts. Ephemeral means its contents are lost when the container restarts. The plan sets its
  size, and when the instance writes more than the plan allows, it restarts and everything in
  ephemeral storage is lost. Move large files into `/home/jovyan` or mounted storage before it
  fills. See
  [Where your files live on an instance](/tir/tir-instance/concepts/where-your-files-live/).

The charts follow CPU and RAM use over time, and on a GPU plan, GPU use, GPU memory, temperature,
and power. Compare **GPU Utilization** with **GPU Memory Used** to tell a GPU that
is busy computing from one that is only holding data in memory.

To look further back, choose **1h**, **1d**, **7d**, or **30d**.

## Read and download the logs

Select **Logs** to see the output of the instance's container, starting with the last 1,000 lines.

- To see a different amount, choose **Last N Lines** or **Last N Seconds** under **Filter By**, and
  set **Value of N**. **None** removes the filter.
- To keep the view current, set **Auto Refresh** to **10 Seconds**, **30 Seconds**, or
  **1 Minute**.
- To keep a copy, click **Download logs**. All the logs are saved to your computer as a JSON file
  named after the instance.

## See the instance's events

Select **Instances Event** to see what has happened to the instance's container recently, and why.

To find out why an instance restarted, look at **Instances Event** and **Logs** while it is
running, and at the **Ephemeral Storage** and **RAM Utilization** figures on **Monitoring**.

## Attach alerts

An alert watches one metric and warns you when it crosses the alert's threshold. On an instance, an
alert can watch CPU use, memory use, workspace disk use, ephemeral storage use, GPU temperature, or
GPU power. An alert on ephemeral storage or memory use warns you before the instance runs out and
is restarted.

1. Open the instance's page and select **Alert**.
2. Choose the alerts under **Select Alerts**, and attach them. To create an alert first, use the
   link on the tab to open the alert form. [Create an alert](/tir/alerts/how-to/create-an-alert/)
   explains its thresholds and notifications.

You can detach an alert from the same tab. Alerts attached to an
instance are removed when you delete the instance.

When a metric points to a problem, such as an instance that keeps restarting, see
[Troubleshooting TIR Instances](/tir/tir-instance/troubleshooting/tir-instance/).