We suspect this happens because the user is linked to a large amount of data. When the profile submission fires, the user-info request attempts to determine the number of records associated with this user in order to display a warning. However, the request takes too long, exceeding 30 seconds in our test case, resulting in a 504 Gateway Timeout.
We have been somewhat conservative with our timeout settings, as they help ensure that unresponsive servers do not leave connections open unnecessarily. However, we may have set them too strictly due to a lack of understanding. Additionally, a different timeout setting may be the actual cause of the issue.
In the CHT-Core nginx config, proxy_read_timeout is set to 3600 (cht-core/nginx/nginx.conf at master · medic/cht-core · GitHub).
I understand you wish to be conservative about timeouts, however large requests can take longer. I suggest you experiment with increasing this timeout. Our safe recommendation is 3600 seconds.
We’re happy to increase the timeout slightly for most routes. However, if there are specific routes that genuinely require an hour, please let us know.
These endpoints should have timeouts above the default. I recommend you experiment with an increased timeout and check whether any of your requests get dropped. We export a metrics endpoint that will give you endpoint stats
: API to interact with CHT Applications | Community Health Toolkit . This can be monitored through watchdog for overtime stats so you can check what your rough timeout should be.
