syncop: Added scaling down logic

RCA:
Whenever the self-heald tests are done with more than 16 replicates
The number of sync procs goes to > 2. These threads never die.

Fix:
Added scaling down logic in syncops so that the threads terminate
themselves whenever the extra thread is idle for ~10 minutes.
Minimum number of threads is still 2.

Tests:
Added logs for launching and terminating procs, made timeout to
6 seconds and ran volume-heal in a while loop. After logs say
max number of procs are launched, attached process to gdb and
verified that the number of syncop threads are 16. Stopped
volume-heal and observed the logs for terminating the procs.
Attached gdb to process again to check that the syncop threads
are just 2. Did this 5 times. Things worked fine. Which procs
were terminated was random. No proc structure was erroneously
re-used. Procs never exceeded 16 and were never < 2.

Change-Id: I61dd9c25cc478ac8cbda190bee841a995b93c55c
BUG: 814074
Signed-off-by: Pranith Kumar K <pranithk@gluster.com>
Reviewed-on: http://review.gluster.org/3195
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

1 files changed, 1 insertions, 0 deletions

diff --git a/libglusterfs/src/syncop.h b/libglusterfs/src/syncop.h
index 619fe7119b9..3aa3c1d01eb 100644
--- a/libglusterfs/src/syncop.h
+++ b/libglusterfs/src/syncop.h
@@ -23,6 +23,7 @@
 
 #define SYNCENV_PROC_MAX 16
 #define SYNCENV_PROC_MIN 2
+#define SYNCPROC_IDLE_TIME 600
 
 struct synctask;
 struct syncproc;