Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1862364

Summary:	Race condition in oslo_privsep daemon
Product:	Red Hat OpenStack	Reporter:	Cédric Jeanneret <cjeanner>
Component:	python-oslo-privsep	Assignee:	mbollo
Status:	CLOSED ERRATA	QA Contact:	nlevinki <nlevinki>
Severity:	high	Docs Contact:
Priority:	high
Version:	16.1 (Train)	CC:	bcafarel, jjoyce, jschluet, lmiccini, mbollo, pkomarov, ralonsoh, slinaber, tvignaud
Target Milestone:	z3	Keywords:	Rebase, Triaged
Target Release:	16.1 (Train on RHEL 8.2)
Hardware:	x86_64
OS:	Linux
Whiteboard:
Fixed In Version:	python-oslo-privsep-1.33.4-1.el8ost	Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2020-12-15 18:36:15 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	1783195

Description Cédric Jeanneret 2020-07-31 08:42:37 UTC

Description of problem:
A race condition might be triggered with the socket used by the daemon.py: since it's provided by an external process, it might happen that external process drops the socket before the end of privsep actions.

This leads to different issues, among them:
- stack trace with "BrokenPipeError: [Errno 32] Broken pipe"
- service hanging while trying to read the socket usint Threading - it will hand there forever and ever


Version-Release number of selected component (if applicable):
python3-oslo-privsep-1.33.3-0.20200310175027.ddde706.el8ost.noarch
python-oslo-privsep-lang-1.33.3-0.20200310175027.ddde706.el8ost.noarch

How reproducible:
Always

Steps to Reproduce:
1. Create a socket for oslo_privsep
2. Start running privsep
3. Drop the socket in the middle of the action

Actual results:
We end with either a BrokenPipe, or a hanging process


Expected results:
It should fail more gracefully, and really NOT hang forever

Additional info:
This was detected while working on the following BZ:
https://bugzilla.redhat.com/show_bug.cgi?id=1783195

Comment 1 Hervé Beraud 2020-08-25 11:41:48 UTC

Hello,

Looks like to an eventlet issue.

Indeed, daemon is an independent process that executes, in privileged mode, the methods passed. Some of those methods make use of libraries like "os" or "threading". Sometimes, those methods are not evenlet safe: the GIL is given to the next user thread and is never returned, that ends in a command timeout.

Unless there is a major reason to monkey patch those libraries, I suggest to revert the monkey patched libraries [1][2][3] when the privileged daemon is forked. That will help to prevent this kind of hanging process and those command timeouts.

[1] https://review.opendev.org/#/c/740970/
[2] https://review.opendev.org/#/c/747906/
[3] https://review.opendev.org/#/c/747907/

Comment 17 errata-xmlrpc 2020-12-15 18:36:15 UTC

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (Red Hat OpenStack Platform 16.1.3 bug fix and enhancement advisory), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHEA-2020:5413