From: Josh Durgin <jdurgin@redhat.com>
Date: Tue, 10 Oct 2017 23:54:31 +0000 (-0400)
Subject: common/options: enable multiple rocksdb compaction threads for filestore
X-Git-Tag: v13.0.1~505^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=023fa810aa6b3af305e9027e3f717e54d1bb2712;p=ceph.git

common/options: enable multiple rocksdb compaction threads for filestore

One of the major benefits of rocksdb over leveldb is multithreaded
compaction. The default of 1 thread does not provide much benefit, and
is insufficient for heavy rgw workloads.

For high-write and delete omap workloads I've seen up to 8 compaction
threads be used.  There's little overhead to having a higher max than
are needed, so set the default to 8.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---

diff --git a/src/common/options.cc b/src/common/options.cc
index fa05f0850df..f3d99913f30 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -3673,7 +3673,7 @@ std::vector<Option> get_global_options() {
     // filestore
 
     Option("filestore_rocksdb_options", Option::TYPE_STR, Option::LEVEL_DEV)
-    .set_default("compaction_readahead_size=2097152")
+    .set_default("max_background_compactions=8;compaction_readahead_size=2097152")
     .set_description(""),
 
     Option("filestore_omap_backend", Option::TYPE_STR, Option::LEVEL_DEV)