Bug 1214649 - BitRot :- File should be removed from signing queue if reopened for writing in 120 second
Summary: BitRot :- File should be removed from signing queue if reopened for writing ...
Keywords:
Status: CLOSED WORKSFORME
Alias: None
Product: GlusterFS
Classification: Community
Component: bitrot
Version: mainline
Hardware: x86_64
OS: Linux
unspecified
low
Target Milestone: ---
Assignee: bugs@gluster.org
QA Contact: Marco Bill-Peter
bugs@gluster.org
URL:
Whiteboard:
: 1214662 (view as bug list)
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2015-04-23 09:59 UTC by Rachana Patel
Modified: 2018-11-13 11:52 UTC (History)
5 users (show)

Fixed In Version:
Clone Of:
: 1224156 (view as bug list)
Environment:
Last Closed: 2018-11-13 11:52:57 UTC
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Embargoed:


Attachments (Terms of Use)

Description Rachana Patel 2015-04-23 09:59:06 UTC
Description of problem:
=======================
BitRot :- When all fds are closed on files, file will be candidate for signing. If file is reopened for writing in next 120 second then it should be removed from queue


Version-Release number of selected component (if applicable):
=============================================================
0.803.gitf64666f.el6.x86_64

How reproducible:
=================
always

Steps to Reproduce:
===================
1. create and mount volume.
2. enable bitrot.
3. create few files.
4. open for write and close it.
5. repeat step 4  for 3-4 times within 5 seconds.

Actual Result:-
===============
File is not being removed from signing queue if re opened within 120 second and signed multiple times

Expect Result:-
===============
File should be removed from signing queue  if reopened for writing in 120 second

Comment 1 Venky Shankar 2015-05-06 04:05:35 UTC
*** Bug 1214662 has been marked as a duplicate of this bug. ***

Comment 2 Raghavendra Bhat 2015-05-08 09:49:28 UTC
This issue has been taken care in the patch sent to handle anonymous fd writes (1207979). http://review.gluster.org/#/c/10233/ takes care of that.

But still there is a small window of multiple notifications on the same gfid being present in the timer wheel if the inode is forgotten in the brick after sending a notification to BitD, but before the notification entry expires from timer-wheel a new write and release happens on that gfid in the brick (now on a newly looked up inode).

This window is pretty negligible in real world use cases. Hence not marking the bug as modified. But its not a point of concern as of now.

Comment 3 Venky Shankar 2015-05-16 08:09:51 UTC
Moving this BZ back to NEW as this scenario is taken care indirectly by the anonfd patch (already merged). As per Comment #2 by Johnny, the probability of the scenario happening is pretty negligible.

Therefore, I'll propose not to track this BZ unless the impact of this bug is experienced more often.

Comment 4 Venky Shankar 2015-05-18 10:08:14 UTC
Rajesh/Anoop,

We're not planning to fix this as of now (read comment #2 and #3 for explanation).

Any concerns?

Comment 7 Amar Tumballi 2018-11-13 11:52:57 UTC
Not seen this issue in last 3 years or so.


Note You need to log in before you can comment on or make changes to this bug.