From what I can see the error occurs when pip requests the index and doesn't receive the response before a 15 second timeout
pip has two ways it tries to get the path to the package
GET /openstack/packagename/
GET /openstack/ (the returned listing is then searched with a non case sensitive search)
If we assume a small number of responses are delayed passed the 15 second timeout then in most cases the problem doesn't manifest itself because one of the two requests gets a hit
Most occurrences of this (if not all) are happening in cases where the subdirectory name and package name don't have matching cases, in this scenario then the first HTTP GET always gets a 404 and we are left with only the get to the main index page (effectively there is no fall-back), so these packages are more likely to see an error due to the 15 second timeout.
How about I submit an Increase to the timeout to see if it improves matters?
export PIP_DEFAULT_TIMEOUT=30
I've been debugging a similar issue in tripleo-ci jobs /bugs.launchpad .net/tripleo/ +bug/1292141
https:/
From what I can see the error occurs when pip requests the index and doesn't receive the response before a 15 second timeout
pip has two ways it tries to get the path to the package packagename/
GET /openstack/
GET /openstack/ (the returned listing is then searched with a non case sensitive search)
If we assume a small number of responses are delayed passed the 15 second timeout then in most cases the problem doesn't manifest itself because one of the two requests gets a hit
Most occurrences of this (if not all) are happening in cases where the subdirectory name and package name don't have matching cases, in this scenario then the first HTTP GET always gets a 404 and we are left with only the get to the main index page (effectively there is no fall-back), so these packages are more likely to see an error due to the 15 second timeout.
How about I submit an Increase to the timeout to see if it improves matters? TIMEOUT= 30
export PIP_DEFAULT_