From c8d343d592de92af91c6f9eecb6aab0ebed54657 Mon Sep 17 00:00:00 2001
From: Patrick Donnelly <pdonnell@redhat.com>
Date: Tue, 9 Jun 2020 16:49:20 -0700
Subject: [PATCH] mds: add maximum random ephemeral pin percentage

This new config is designed to prevent creating too many subtrees from a
random ephemeral pinning policy.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
---
 src/common/options.cc | 7 +++++++
 src/mds/CInode.cc     | 3 ++-
 src/mds/MDCache.cc    | 4 ++++
 src/mds/MDCache.h     | 2 ++
 src/mds/MDSRank.cc    | 1 +
 src/mds/Server.cc     | 8 ++++++++
 6 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/src/common/options.cc b/src/common/options.cc
index 5b232c7f88e6..55e4d5539c92 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -8005,6 +8005,13 @@ std::vector<Option> get_mds_options() {
     .set_description("allow ephemeral random pinning of the loaded subtrees")
     .set_long_description("probabilistically pin the loaded directory inode and the subtree beneath it to an MDS based on the consistent hash of the inode number. The higher this value the more likely the loaded subtrees get pinned"),
 
+    Option("mds_export_ephemeral_random_max", Option::TYPE_FLOAT, Option::LEVEL_ADVANCED)
+    .set_default(0.01)
+    .set_flag(Option::FLAG_RUNTIME)
+    .set_description("the maximum percent permitted for random ephemeral pin policy")
+    .set_min_max(0.0, 1.0)
+    .add_see_also("mds_export_ephemeral_random"),
+
     Option("mds_export_ephemeral_distributed", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
     .set_default(false)
     .set_flag(Option::FLAG_RUNTIME)
diff --git a/src/mds/CInode.cc b/src/mds/CInode.cc
index 39239f8fcd9a..03ad469d5071 100644
--- a/src/mds/CInode.cc
+++ b/src/mds/CInode.cc
@@ -5514,6 +5514,7 @@ double CInode::get_ephemeral_rand(bool inherit) const
    * have a parent yet.
    */
   const CInode *in = this;
+  double max = mdcache->export_ephemeral_random_max;
   while (true) {
     if (in->is_system())
       break;
@@ -5525,7 +5526,7 @@ double CInode::get_ephemeral_rand(bool inherit) const
       break;
 
     if (in->get_inode().export_ephemeral_random_pin > 0.0)
-      return in->get_inode().export_ephemeral_random_pin;
+      return std::min(in->get_inode().export_ephemeral_random_pin, max);
 
     /* An export_pin overrides only if no closer parent (incl. this one) has a
      * random pin set.
diff --git a/src/mds/MDCache.cc b/src/mds/MDCache.cc
index 58ce50128fbe..46974b7024d6 100644
--- a/src/mds/MDCache.cc
+++ b/src/mds/MDCache.cc
@@ -155,6 +155,7 @@ MDCache::MDCache(MDSRank *m, PurgeQueue &purge_queue_) :
 
   export_ephemeral_distributed_config =  g_conf().get_val<bool>("mds_export_ephemeral_distributed");
   export_ephemeral_random_config =  g_conf().get_val<bool>("mds_export_ephemeral_random");
+  export_ephemeral_random_max = g_conf().get_val<double>("mds_export_ephemeral_random_max");
 
   lru.lru_set_midpoint(g_conf().get_val<double>("mds_cache_mid"));
 
@@ -243,6 +244,9 @@ void MDCache::handle_conf_change(const std::set<std::string>& changed, const MDS
     }
     mds->balancer->handle_export_pins();
   }
+  if (changed.count("mds_export_ephemeral_random_max")) {
+    export_ephemeral_random_max = g_conf().get_val<double>("mds_export_ephemeral_random_max");
+  }
   if (changed.count("mds_health_cache_threshold"))
     cache_health_threshold = g_conf().get_val<double>("mds_health_cache_threshold");
   if (changed.count("mds_cache_mid"))
diff --git a/src/mds/MDCache.h b/src/mds/MDCache.h
index 1cce2ac33b51..11fe5b9e3a6b 100644
--- a/src/mds/MDCache.h
+++ b/src/mds/MDCache.h
@@ -1001,6 +1001,8 @@ class MDCache {
 
   OpenFileTable open_file_table;
 
+  double export_ephemeral_random_max = 0.0;
+
  protected:
   // track master requests whose slaves haven't acknowledged commit
   struct umaster {
diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
index fed587d5d33e..73527884f5fb 100644
--- a/src/mds/MDSRank.cc
+++ b/src/mds/MDSRank.cc
@@ -3600,6 +3600,7 @@ const char** MDSRankDispatcher::get_tracked_conf_keys() const
     "mds_dump_cache_threshold_formatter",
     "mds_enable_op_tracker",
     "mds_export_ephemeral_random",
+    "mds_export_ephemeral_random_max",
     "mds_export_ephemeral_distributed",
     "mds_health_cache_threshold",
     "mds_inject_migrator_session_race",
diff --git a/src/mds/Server.cc b/src/mds/Server.cc
index a261fc4c3110..1b8a2ac03110 100644
--- a/src/mds/Server.cc
+++ b/src/mds/Server.cc
@@ -5675,6 +5675,14 @@ void Server::handle_set_vxattr(MDRequestRef& mdr, CInode *cur)
       return;
     }
 
+    if (val < 0.0 || 1.0 < val) {
+      respond_to_request(mdr, -EDOM);
+      return;
+    } else if (mdcache->export_ephemeral_random_max < val) {
+      respond_to_request(mdr, -EINVAL);
+      return;
+    }
+
     if (!xlock_policylock(mdr, cur))
       return;
 
-- 
2.47.3