Skip to content
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

mds: don't maintain bloom filters in standby replay #12133

Merged
merged 2 commits into from Dec 7, 2016
Merged

Conversation

jcsp
Copy link
Contributor

@jcsp jcsp commented Nov 22, 2016

@jcsp jcsp added bug-fix cephfs Ceph File System labels Nov 22, 2016
Copy link
Member

@gregsfortytwo gregsfortytwo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Couple questions/changes.

fs=0
for name in a b c d e f g h i j k l m n o p
do
sleep 5 # wait for daemons to make it into FSMap
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This appears to be sleeping 5 seconds for each fs. I don't think that was meant to happen?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed

CInode *in = dn->get_linkage()->get_inode();

if (dn->get_linkage()->is_primary()) {
in->get_nested_dirfrags(subdirs);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I must be misremembering how EExport works. Why do we need to go down one level to pull out the child CDir bloom filters, but not all the way to the leaves?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

remove_blooms calls itself, it does go down to the leaves.

@ukernel
Copy link
Contributor

ukernel commented Nov 23, 2016

I think it's better to avoid adding bloom filter while standby replay

John Spray added 2 commits November 23, 2016 14:17
The daemons will get assigned to roles in the
order they start otherwise.  The trick is
to let them all go into the FSMap first and
then increase max_mds.

Signed-off-by: John Spray <john.spray@redhat.com>
...and thereby avoid needing to clear them when
replaying an EExport

Fixes: http://tracker.ceph.com/issues/16924
Signed-off-by: John Spray <john.spray@redhat.com>
@jcsp jcsp changed the title mds: clear bloom filters on CDir when replaying EExport mds: don't maintain bloom filters in standby replay Nov 23, 2016
@jcsp
Copy link
Contributor Author

jcsp commented Nov 23, 2016

@ukernel good point, I agree. I've updated this.

@gregsfortytwo
Copy link
Member

Reviewed-by: Greg Farnum gfarnum@redhat.com

@jcsp jcsp merged commit 4ed4224 into ceph:master Dec 7, 2016
@jcsp jcsp deleted the wip-16924 branch December 7, 2016 14:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
3 participants