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

os/bluestore/BlueFS: tune flushing of writes #13032

Merged
merged 3 commits into from Jan 25, 2017

Conversation

Projects
None yet
2 participants
@liewegas
Member

liewegas commented Jan 20, 2017

No description provided.

@liewegas liewegas added the bluestore label Jan 20, 2017

@liewegas

This comment has been minimized.

Show comment
Hide comment
@liewegas

liewegas Jan 20, 2017

Member

@rchagam This may help a bit as well...

Member

liewegas commented Jan 20, 2017

@rchagam This may help a bit as well...

@rchagam

This comment has been minimized.

Show comment
Hide comment
@rchagam

rchagam Jan 23, 2017

Contributor

Tested this patch yesterday - runs into assertion failure in BlueFS.cc
**
/root/ceph/src/os/bluestore/BlueFS.cc: 1578: FAILED assert(last.unused_tail_length() == zlen)
**
Detailed stack:
one.txt

Contributor

rchagam commented Jan 23, 2017

Tested this patch yesterday - runs into assertion failure in BlueFS.cc
**
/root/ceph/src/os/bluestore/BlueFS.cc: 1578: FAILED assert(last.unused_tail_length() == zlen)
**
Detailed stack:
one.txt

liewegas added some commits Jan 20, 2017

os/bluestore/BlueFS: preallocate larger write buffer
Signed-off-by: Sage Weil <sage@redhat.com>
os/bluestore/BlueFS: increase size threshold before we flush (and gen…
…erate io)

Having this too high means you might be more bursty. In practice,
though, the commit path is doing explicit syncs on small chunks
anyway.  And compaction work should probably stay reasonably chunky.

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

This comment has been minimized.

Show comment
Hide comment
@liewegas

liewegas Jan 23, 2017

Member

thanks, fixing

Member

liewegas commented Jan 23, 2017

thanks, fixing

os/bluestore/BlueFS: tolerate longer tail
The important requirement here is that we have enough buffer to write the
whole page.

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

This comment has been minimized.

Show comment
Hide comment
@liewegas
Member

liewegas commented Jan 23, 2017

@rchagam fixed

@rchagam

This comment has been minimized.

Show comment
Hide comment
@rchagam

rchagam Jan 23, 2017

Contributor

@liewegas works!! looking at performance analysis - will send you update later today.

Contributor

rchagam commented Jan 23, 2017

@liewegas works!! looking at performance analysis - will send you update later today.

@rchagam

This comment has been minimized.

Show comment
Hide comment
@rchagam

rchagam Jan 24, 2017

Contributor

@liewegas tested with rados bench 4k writes (10 min run) - slightly better throughput and latency with tuned write flushes:
bluestore_rtc - 10.116 (ms), 7,590,743 writes
bluestore_rtc + bluefs tuned flush writes - 10.042 (ms), 7,646,524 writes

Contributor

rchagam commented Jan 24, 2017

@liewegas tested with rados bench 4k writes (10 min run) - slightly better throughput and latency with tuned write flushes:
bluestore_rtc - 10.116 (ms), 7,590,743 writes
bluestore_rtc + bluefs tuned flush writes - 10.042 (ms), 7,646,524 writes

@liewegas liewegas merged commit 3c59280 into ceph:master Jan 25, 2017

3 checks passed

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

@liewegas liewegas deleted the liewegas:wip-bluefs-append branch Jan 25, 2017

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