From b076a947fcc3cc0b99a8591a240ee29322d7198a Mon Sep 17 00:00:00 2001
From: David Anderson <davea@ssl.berkeley.edu>
Date: Sun, 9 Mar 2014 17:09:21 -0700
Subject: [PATCH] client: work fetch tweak to avoid starvation in a particular
 case

My commit of Feb 7 caused work fetch to project P
to be deferred for up to 5 min if an upload to P is active,
even if some instances are idle.
This was to deal with a case where the idleness was caused
by a jobs-in-progress limit by P,
and work requests lead to long backoff.

However, this can cause instances to be idle unnecessarily.
I changed things so that, if instances are idle,
a work fetch can happen even during upload.
But only one such fetch will be done.
---
 client/cs_apps.cpp      |  4 ++++
 client/cs_scheduler.cpp | 10 ++++++++--
 client/work_fetch.cpp   |  1 +
 client/work_fetch.h     |  4 ++++
 4 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/client/cs_apps.cpp b/client/cs_apps.cpp
index b107c53f65..88f0d4a3c8 100644
--- a/client/cs_apps.cpp
+++ b/client/cs_apps.cpp
@@ -205,6 +205,10 @@ int CLIENT_STATE::app_finished(ACTIVE_TASK& at) {
     double elapsed_time = now - rec_interval_start;
     work_fetch.accumulate_inst_sec(&at, elapsed_time);
 
+    rp->project->pwf.request_if_idle_and_uploading = true;
+        // set this to allow work fetch if idle instance,
+        // even before upload finishes
+
     return 0;
 }
 
diff --git a/client/cs_scheduler.cpp b/client/cs_scheduler.cpp
index 69313350db..d016b14437 100644
--- a/client/cs_scheduler.cpp
+++ b/client/cs_scheduler.cpp
@@ -534,7 +534,13 @@ bool CLIENT_STATE::scheduler_rpc_poll() {
     p = work_fetch.choose_project();
     if (p) {
         if (actively_uploading(p)) {
-            //if (!idle_request()) {
+            bool dont_request = true;
+            if (p->pwf.request_if_idle_and_uploading) {
+                if (idle_request()) {
+                    dont_request = false;
+                }
+            }
+            if (dont_request) {
                 if (log_flags.work_fetch_debug) {
                     msg_printf(p, MSG_INFO,
                         "[work_fetch] deferring work fetch; upload active"
@@ -542,7 +548,7 @@ bool CLIENT_STATE::scheduler_rpc_poll() {
                 }
                 p->sched_rpc_pending = 0;
                 return false;
-            //}
+            }
         }
         scheduler_op->init_op_project(p, RPC_REASON_NEED_WORK);
         return true;
diff --git a/client/work_fetch.cpp b/client/work_fetch.cpp
index fa21178ad2..08843dfa17 100644
--- a/client/work_fetch.cpp
+++ b/client/work_fetch.cpp
@@ -955,6 +955,7 @@ void WORK_FETCH::handle_reply(
             p->rsc_pwf[i].clear_backoff();
         }
     }
+    p->pwf.request_if_idle_and_uploading = false;
 }
 
 // set up for initial RPC.
diff --git a/client/work_fetch.h b/client/work_fetch.h
index 75f7d09f29..b58a77193a 100644
--- a/client/work_fetch.h
+++ b/client/work_fetch.h
@@ -274,6 +274,10 @@ struct PROJECT_WORK_FETCH {
     int cant_fetch_work_reason;
     int compute_cant_fetch_work_reason(PROJECT*);
     int n_runnable_jobs;
+    bool request_if_idle_and_uploading;
+        // Set when a job finishes.
+        // If we're uploading but a resource is idle, make a work request.
+        // If this succeeds, clear the flag.
     PROJECT_WORK_FETCH() {
         memset(this, 0, sizeof(*this));
     }