gluster-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gluster-devel] [RFC] core: Introduce internal fop types.


From: Vijay Bellur
Subject: Re: [Gluster-devel] [RFC] core: Introduce internal fop types.
Date: Thu, 02 Jan 2014 19:41:58 +0530
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.1.0

On 12/13/2013 10:09 AM, Raghavendra Gowdappa wrote:
Hi all,

It is increasingly becoming a necessity that xlators other than dht/afr to be 
aware of internal fops (fops not sent by kernel). Some of the xlators that use 
internal fops are:

1. changelog - to ignore link/unlink (triggered as part of rename) on 
dht-linkfiles
2. posix -  to overload mknod call to act as a link if file already exists 
(consumers - afr and gfid-acess)

We are seeing similar needs in quota and marker:

1. Quota-enforcer: To treat link/unlinks on dht-datafiles triggered as part of 
rename equivalent to renames [1]
2. Marker: To not account dht-linkfiles

Couple of months back when working on quota, there was a consensus that 
GLUSTERFS_INTERNAL_FOP_KEY is overloaded and should be cleaned up. Hence, I am 
proposing to introduce internal-fop types [2]. Please let me know your opinions.

[1] http://review.gluster.com/#/c/6399/
[2] http://review.gluster.org/#/c/6497/


I think having more apt keys for performing relevant actions is good to have. While this change is performed, we need to ensure that the behavior of posix/changelog other consumers remain backward compatible.

-Vijay




reply via email to

[Prev in Thread] Current Thread [Next in Thread]