draw.io for Confluence has an option in the plugin admin screen to enable image and PDF generation via an external server. The plugin admin screen is available by navigating to the plugins management screen, selecting draw.io then selecting the "configure" button in the listing.
The reason you might do this is text in image generation is dependent on what fonts are installed on the server that Confluence is running on. The font support on Linux is relatively weak and substituted fonts can often be a poor match. In additional, many unicode characters are not supported by many standard font families in Linux. Windows based servers can suffer from this problem to a lesser extent.
Image generation is used both within the draw.io editor and in the Confluence page that holds the diagram(s). The editor has export options to PNG and PDF. In page edit mode a rendered PNG is used as a placeholder. This placeholder is also used in various page export options, such as PDF export.
draw.io does perform client-side image generation on modern Chrome, Firefox and Microsoft Edge browsers, but is unable to for Internet Explorer and Safari. If you have users on IE and Safari you have the option to allow those users to generate images via our external image generation server by ticking the option in the plugin admin screen.
The external image servers are entirely controlled by us and located in secure US data centres. When generating images the image data is sent securely to the server, the image generated and returned, and the data deleted entirely from the server (including any diagram data in logs). No diagram data is retained or transmitted from the servers. Nobody except for 2 senior JGraph engineers have access to the servers and they are not permitted to move any user data off of those servers.