Bug 1392135 - Possible data corruption when in-memory cache is enabled
Summary: Possible data corruption when in-memory cache is enabled
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Red Hat Ceph Storage
Classification: Red Hat
Component: RBD
Version: 1.3.3
Hardware: Unspecified
OS: Unspecified
unspecified
urgent
Target Milestone: rc
: 1.3.4
Assignee: Jason Dillaman
QA Contact: ceph-qe-bugs
URL:
Whiteboard:
Depends On:
Blocks: 1392136
TreeView+ depends on / blocked
 
Reported: 2016-11-05 00:44 UTC by Jason Dillaman
Modified: 2018-03-07 23:52 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
: 1392136 (view as bug list)
Environment:
Last Closed: 2018-03-07 23:52:26 UTC
Target Upstream Version:

Attachments (Terms of Use)
Add an attachment (proposed patch, testcase, etc.)


Links
System ID Private Priority Status Summary Last Updated
Ceph Project Bug Tracker 16546 0 None None None 2016-11-05 00:44:29 UTC

Description Jason Dillaman 2016-11-05 00:44:29 UTC 
Description of problem:
With writeback cache enabled, there is a possibility that read requests serviced through the cache will be corrupted. 

Version-Release number of selected component (if applicable):
1.2, 1.3

How reproducible:
Reports indicate nearly 100% given enough time and high IOPS workload (i.e. a database load simulator).

Steps to Reproduce:
1. Reported to fail with Windows Server 2012R2 and SQL Server backed by RBD image with RBD writeback cache enabled.
2. Use SQLioSim to inject load into the database

Actual results:
Expected FileId: 0x0
Received FileId: 0x0
Expected PageId: 0xCB19C
Received PageId: 0xCB19A (does not match expected)
Received CheckSum: 0x9F444071
Calculated CheckSum: 0x89603EC9 (does not match expected)
Received Buffer Length: 0x2000

Expected results:
No IO errors

Additional info:
Comment 3 Jason Dillaman 2016-11-05 00:48:24 UTC 
Upstream, backport PR against hammer branch: https://github.com/ceph/ceph/pull/11618
Note You need to log in before you can comment on or make changes to this bug.