From 1a08a4126650f0ff82187124a27b0b9efcfc7726 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@redhat.com>
Date: Mon, 9 Apr 2018 07:20:05 -0500
Subject: [PATCH] mon/OSDMonitor: add mon_osd_max_initial_pgs to cap initial
 pool pgs

Configure how many initial PGs we create a pool with.  If the user wants
more than this then we do subsequent splits.

Default to 1024, so that pool creation works in the usual way for most users,
but does some splitting for very large pools/clusters.

Signed-off-by: Sage Weil <sage@redhat.com>
---
 src/common/options.cc | 7 ++++++-
 src/mon/OSDMonitor.cc | 5 ++++-
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/src/common/options.cc b/src/common/options.cc
index 938a942724481..595497fc4269f 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -1261,7 +1261,12 @@ std::vector<Option> get_global_options() {
 
     Option("mon_osd_max_creating_pgs", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(1024)
-    .set_description(""),
+    .set_description("Maximum number of PGs the mon will create at once"),
+
+    Option("mon_osd_max_initial_pgs", Option::TYPE_INT, Option::LEVEL_ADVANCED)
+    .set_default(1024)
+    .set_description("Maximum number of PGs a pool will created with")
+    .set_long_description("If the user specifies more PGs than this, the cluster will subsequently split PGs after the pool is created in order to reach the target."),
 
     Option("mon_tick_interval", Option::TYPE_INT, Option::LEVEL_ADVANCED)
     .set_default(5)
diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index aa1c5a4d01519..ce47155e78e5f 100644
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -6692,7 +6692,10 @@ int OSDMonitor::prepare_new_pool(string& name,
   pi->crush_rule = crush_rule;
   pi->expected_num_objects = expected_num_objects;
   pi->object_hash = CEPH_STR_HASH_RJENKINS;
-  pi->set_pg_num(1);
+  auto max = g_conf().get_val<int64_t>("mon_osd_max_initial_pgs");
+  pi->set_pg_num(
+    max > 0 ? std::min<uint64_t>(pg_num, std::max<int64_t>(1, max))
+    : pg_num);
   pi->set_pg_num_pending(pi->get_pg_num(), pending_inc.epoch);
   pi->set_pg_num_target(pg_num);
   pi->set_pgp_num(pi->get_pg_num());
-- 
2.39.5