Memory leak when refreshing


I work in a bank and I’m testing Grafana to display market data and logs, to know if it will be a good solution to replace our current system.

I have an issue: the browser displaying the dashboard is suffering memory leak. Memory consumption grows constantly, from 100 MB at start until it reaches 1.5 GB three hours later. Then the browser crashes.

I’m using Grafana 4.5.2 with Influxdb 1.3.5-1 and Elasticsearch 5.5.2. I tried different browsers (chromium, firefox, opera, IE, always latest version) and I tried to run Grafana on windows and on Linux, on 3 different computers. I tested running locally or using the internal network of my company.
The problem is always the same.

My dashboard is auto-refreshing every second. When I deactivate auto-refreshing, the memory leak stops. If refreshing is set to a lower frequency, the leak is slower. So it seems to be caused by refreshing the page.

I tried to deactivate some parts of the dashboard: for example leaving only charts and tables connected to the influxDb source, of leaving only the Elasticsearch source. I tried also to deactivate only tables, or only charts. This slows the leak but do not stop it.

I tried also with a completely different dashboard, connected to a different data source: memory leak is still there, as long as I’m refreshing often (1s or 5s).

Here is for example a screenshot of chromium task manager showing memory consumption after 1 hour. The memory consumption is at 600 MB and continues to grow:

I tried to look at other topics on github, some users were also having a similar issue one year ago:

Upgrading Grafana seemed to solve these users issue, but unfortunately not for me because I’m already using the latest version.

Could you please help me with this issue?



what panels do you use? any template variables?

I’m using graph and table panels.

I have 3 template variables, in query form, like:

what are the option for the template variables? refresh on dashboard load or time range change?

It’s refresh on dashboard load.

However I made some other tests and I found what is causing the leak: it’s my company proxy script. When I disable it, the leak disappears.
I am unfortunately not authorized to disable this script for the users so it will be an issue for deploying Grafana within the company.

Thanks for your time though.