Keystone appears to initiate each new request using the previous' request-id
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Identity (keystone) |
Invalid
|
Undecided
|
Unassigned |
Bug Description
Environment
~~~~~~~~~~~
* openstack-ansible version: 16.0.2
* Target OS: Ubuntu 16.04 Xenial
* Keystone deployed in containers, running on uWSGI (per OSA defaults)
* Keystone baseline (as provided by OSA): 6a67918f9d5f395
Symptom
~~~~~~~
When running Keystone with debug=True, one can observe in keystone.log that each incoming request appears to "borrow" the req Id from the previous one served by that particular uWSGI process.
Analysis
~~~~~~~~
This may be just cosmetic but one wonders if this is the indication of smth executing under the wrong context (and hence could have security implications?)
Evidence
~~~~~~~~
In this slightly edited log excerpt from a specific worker (11207): http://
Notes:
1. there's a ~5s delay before 20:30:25.957 which is a different issue (actually the one I was investigating which lead me to notice the pattern reported here
2. The log has been slightly edited: I removed the "UUID1 UUID2 default - default" suffix after the req Id for legibility (both UUIDs were constants); this is what allowed me to spot the issue.
description: | updated |
Changed in keystone: | |
milestone: | none → queens-rc1 |
milestone: | queens-rc1 → none |
I can't quite reproduce this with devstack but I'm seeing a related strange issue where there's no request ID at all for the "Authenticating user token" log: http:// paste.openstack .org/show/ 652190/
Not sure if either of these are caused by keystone or by oslo.middleware. Could you share your keystone-paste.ini file?