(PE-37376) Initialize one JRuby instance first when initializing pool

Puppet may apply a "settings catalog" when it initializes. This will
check the filesystem and create any directories specified by settings
that do not already exist. This will cause errors if Puppet does need to
create directories and is initialized in parallel. To resolve this when
priming the pool we now instantiate one instance first before
initializing the remaining instances in parallel.

Author: justinstoller

src/clj/puppetlabs/services/jruby_pool_manager/impl/jruby_agents.clj

@@ -70,25 +70,29 @@
 
 (schema/defn ^:always-validate
   prime-pool!
-  "Sequentially fill the pool with new JRubyInstances.  NOTE: this
-  function should never be called except by the modify-instance-agent
-  to create a pool's initial jruby instances."
+  "Fill the pool with new JRubyInstances.  Instantiates the first JRuby (Puppet
+  will sometimes alter the filesystem on first instantiation) and the remaining
+  instances in parallel.  NOTE: this function should never be called except by
+  the modify-instance-agent to create a pool's initial jruby instances."
   [{:keys [config] :as pool-context} :- jruby-schemas/PoolContext]
   (log/debug (format "%s\n%s"
                      (i18n/trs "Initializing JRubyInstances with the following settings:")
                      (ks/pprint-to-string config)))
   (let [pool (jruby-internal/get-pool pool-context)
         creation-service (jruby-internal/get-creation-service pool-context)
         total (.remainingCapacity pool)
-        ids (->> total range (map inc))
+        [first-id & ids] (->> total range (map inc))
         add-instance* (fn [id]
                           (log/debug (i18n/trs "Priming JRubyInstance {0} of {1}"
                                                id count))
                           (add-instance pool-context id)
                           (log/info (i18n/trs "Finished creating JRubyInstance {0} of {1}"
                                               id count)))
+        initial-task (fn [] (add-instance* first-id))
         tasks (for [id ids] (fn [] (add-instance* id)))]
-    (execute-tasks! tasks creation-service)))
+    (execute-tasks! [initial-task] creation-service)
+    (when (seq ids)
+      (execute-tasks! tasks creation-service))))
 
 (schema/defn ^:always-validate
   flush-instance!