summaryrefslogtreecommitdiffstats
path: root/src/ceph/doc/dev/osd_internals/wbthrottle.rst
diff options
context:
space:
mode:
authorQiaowei Ren <qiaowei.ren@intel.com>2018-01-04 13:43:33 +0800
committerQiaowei Ren <qiaowei.ren@intel.com>2018-01-05 11:59:39 +0800
commit812ff6ca9fcd3e629e49d4328905f33eee8ca3f5 (patch)
tree04ece7b4da00d9d2f98093774594f4057ae561d4 /src/ceph/doc/dev/osd_internals/wbthrottle.rst
parent15280273faafb77777eab341909a3f495cf248d9 (diff)
initial code repo
This patch creates initial code repo. For ceph, luminous stable release will be used for base code, and next changes and optimization for ceph will be added to it. For opensds, currently any changes can be upstreamed into original opensds repo (https://github.com/opensds/opensds), and so stor4nfv will directly clone opensds code to deploy stor4nfv environment. And the scripts for deployment based on ceph and opensds will be put into 'ci' directory. Change-Id: I46a32218884c75dda2936337604ff03c554648e4 Signed-off-by: Qiaowei Ren <qiaowei.ren@intel.com>
Diffstat (limited to 'src/ceph/doc/dev/osd_internals/wbthrottle.rst')
-rw-r--r--src/ceph/doc/dev/osd_internals/wbthrottle.rst28
1 files changed, 28 insertions, 0 deletions
diff --git a/src/ceph/doc/dev/osd_internals/wbthrottle.rst b/src/ceph/doc/dev/osd_internals/wbthrottle.rst
new file mode 100644
index 0000000..a3ae00d
--- /dev/null
+++ b/src/ceph/doc/dev/osd_internals/wbthrottle.rst
@@ -0,0 +1,28 @@
+==================
+Writeback Throttle
+==================
+
+Previously, the filestore had a problem when handling large numbers of
+small ios. We throttle dirty data implicitely via the journal, but
+a large number of inodes can be dirtied without filling the journal
+resulting in a very long sync time when the sync finally does happen.
+The flusher was not an adequate solution to this problem since it
+forced writeback of small writes too eagerly killing performance.
+
+WBThrottle tracks unflushed io per hobject_t and ::fsyncs in lru
+order once the start_flusher threshold is exceeded for any of
+dirty bytes, dirty ios, or dirty inodes. While any of these exceed
+the hard_limit, we block on throttle() in _do_op.
+
+See src/os/WBThrottle.h, src/osd/WBThrottle.cc
+
+To track the open FDs through the writeback process, there is now an
+fdcache to cache open fds. lfn_open now returns a cached FDRef which
+implicitely closes the fd once all references have expired.
+
+Filestore syncs have a sideeffect of flushing all outstanding objects
+in the wbthrottle.
+
+lfn_unlink clears the cached FDRef and wbthrottle entries for the
+unlinked object when the last link is removed and asserts that all
+outstanding FDRefs for that object are dead.