[GH-ISSUE #1901] Consider changing `-o noobj_cache` default #964

New issue

Closed

opened 2026-03-04 01:50:17 +03:00 by kerem · 2 comments

kerem commented

2026-03-04 01:50:17 +03:00

Owner

Originally created by @gaul on GitHub (Feb 20, 2022).
Original GitHub issue: https://github.com/s3fs-fuse/s3fs-fuse/issues/1901

Currently the noobj_cache is disabled but we should consider enabling it by default, similar to #927. I am not sure if we need a mechanism to disable it?

Originally created by @gaul on GitHub (Feb 20, 2022). Original GitHub issue: https://github.com/s3fs-fuse/s3fs-fuse/issues/1901 Currently the `noobj_cache` is disabled but we should consider enabling it by default, similar to #927. I am not sure if we need a mechanism to disable it?

kerem

2026-03-04 01:50:17 +03:00

closed this issue
added the
performance
label

kerem commented

2026-03-04 01:50:18 +03:00

Author

Owner

@ggtakec commented on GitHub (Feb 20, 2022):

I agree with enabling it.
But, if we enable the default value, there may be some user(how many users?) confusion.
(It is Imagined that the degree of influence is proportional to the default value of the cache expiration time.)
I think we need to know how many users would be affected ...

In order to disable it, we need to prepare a dedicated option(ex. disable_noobj_cache) and so on.
(Now we have created option names when it is needed, there is no uniformity in option names. So I would like to organize the options as a whole, including the option names.)

@ggtakec commented on GitHub (Feb 20, 2022): I agree with enabling it. But, if we enable the default value, there may be some user(how many users?) confusion. (It is Imagined that the degree of influence is proportional to the default value of the cache expiration time.) I think we need to know how many users would be affected ... In order to disable it, we need to prepare a dedicated option(ex. `disable_noobj_cache`) and so on. (Now we have created option names when it is needed, there is no uniformity in option names. So I would like to organize the options as a whole, including the option names.)

kerem commented

2026-03-04 01:50:18 +03:00

Author

Owner

@gaul commented on GitHub (Feb 20, 2022):

6b78bfdf4b added this flag in 2013 but does not discuss the default behavior. I think that if we can understand the impact on users via tests like #1900 (and possibly more) then we can have confidence to change the default. Agree that adding a flag to restore the current behavior is a safe idea. Probably better to change these kinds of defaults soon after a release to give the most testing time.

@gaul commented on GitHub (Feb 20, 2022): 6b78bfdf4b504629aa36d127ed1b3dbd6d44312b added this flag in 2013 but does not discuss the default behavior. I think that if we can understand the impact on users via tests like #1900 (and possibly more) then we can have confidence to change the default. Agree that adding a flag to restore the current behavior is a safe idea. Probably better to change these kinds of defaults soon after a release to give the most testing time.

kerem referenced this issue

2026-03-04 02:01:54 +03:00

[PR #967] [MERGED] Increase FdEntity reference count when returning #1733