New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

osd: reduce rados_max_object_size from 100 GB -> 128 MB #15520

Merged
merged 1 commit into from Jun 7, 2017

Conversation

Projects
None yet
6 participants
@liewegas
Member

liewegas commented Jun 6, 2017

I'm not sure why we had this so huge! 128 MB is a pretty reasonable size.

qa

  • rados
@xiexingguo

Yay! This sounds better.

@gregsfortytwo

This comment has been minimized.

Member

gregsfortytwo commented Jun 7, 2017

We should be cautious about this. We already (separately) limit the size of a write to 100MB, which will stop naive users from doing something inappropriate, and I seem to recall 128MB is the default Hadoop chunk these days. I think some (admittedly odd) users have wanted sparse objects of small size but very large offsets, which this would prevent.
Why so small an object for the default? (I'm not sure about deep scrub, but recovery works in sub-chunks of separately-configured size and ought to handle much larger ones just fine).

@liewegas

This comment has been minimized.

Member

liewegas commented Jun 7, 2017

Recovery and scrub do small IOs, but they block requests at object granularity, so it still interrupts service. This is similar to why big rgw index objects are so problematic. :(

@liewegas liewegas requested review from jdurgin, yehudasa and jcsp Jun 7, 2017

@jdurgin

This comment has been minimized.

Member

jdurgin commented Jun 7, 2017

This is a default. It can always be changed for crazy use cases. I'm happy with 128MB. Even 64 or 32 would likely be fine for the vast majority of users.

@jdurgin

jdurgin approved these changes Jun 7, 2017

@jcsp

This comment has been minimized.

Contributor

jcsp commented Jun 7, 2017

Mentioning this in the cephfs docs too #15550 for benefit of anyone who tries to set their stripe size too high and hits this limit.

@jcsp

jcsp approved these changes Jun 7, 2017

@gregsfortytwo

This comment has been minimized.

Member

gregsfortytwo commented Jun 7, 2017

Fair enough, I'm on board. :)

@yehudasa

We can have this in place until we figure out scrubs and recovery. But we do need to prioritize that.

osd: reduce rados_max_object_size from 100 GB -> 128 MB
I'm not sure why we had this so huge!  128 MB is a pretty reasonable size.

Signed-off-by: Sage Weil <sage@redhat.com>

@liewegas liewegas merged commit 8a0d672 into ceph:master Jun 7, 2017

2 of 3 checks passed

default Build started sha1 is merged.
Details
Signed-off-by all commits in this PR are signed
Details
Unmodifed Submodules submodules for project are unmodified
Details

@liewegas liewegas deleted the liewegas:wip-max-object-size branch Jun 7, 2017

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment