Bug 1685023

Summary: FD processes for larger files are not closed soon after FOP finished
Product: [Community] GlusterFS Reporter: david.spisla
Component: coreAssignee: Kotresh HR <khiremat>
Status: CLOSED UPSTREAM QA Contact:
Severity: high Docs Contact:
Priority: medium    
Version: mainlineCC: bugs, pasik, rabhat
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2020-03-12 12:45:09 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:

Description david.spisla 2019-03-04 07:51:58 UTC
Description of problem:
I did some observations concerning the bitrot daemon. It seems to be that the bitrot signer is signing files depending on file size. I copied files with different sizes into a volume and I was wonderung because the files get their signature not the same time (I keep the expiry time default with 120). Here are some examples:

300 KB file ~2-3 m 
70 MB file ~ 40 m 
115 MB file ~ 1 Sh
800 MB file ~ 4,5 h

There was already a bug from 2016:
https://bugzilla.redhat.com/show_bug.cgi?id=1378466

I also figured out this discussion: https://lists.gluster.org/pipermail/gluster-users/2016-September/028354.html
Kotresh mentioned there that the problem is because for some files, fd process are still up in the brick process list. Bitrot signer can only sign a file if the fd is closed. And according to my observations it seems to be that as bigger a file is as longer the fd is still up. I could verify this with a 500MiB file and some smaller files. After a specific time only the fd for the 500MiB was up and the file still had no signature, for the smaller files there were no fds and they already had a signature.

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

Gluster v5.3

Actual behaviour: 
The fd processes are still up for a specific time (maybe depend on file size???) after FOP finished and bitd can't sign the file

Expected behaviour:
The fd processes should be closed soon after FOP finished so that bitd can sign the file

Comment 1 Worker Ant 2020-03-12 12:45:09 UTC
This bug is moved to https://github.com/gluster/glusterfs/issues/938, and will be tracked there from now on. Visit GitHub issues URL for further details