cb8e9e
From 8dce156a30800bc404679d380b2ab75bd680dd10 Mon Sep 17 00:00:00 2001
cb8e9e
From: Gaurav Kumar Garg <ggarg@redhat.com>
cb8e9e
Date: Fri, 11 Sep 2015 17:55:42 +0530
cb8e9e
Subject: [PATCH 327/330] glusterd: disable ping timer b/w glusterd and make epoll thread count to 1
cb8e9e
cb8e9e
Currently glusterd is crashing when enable/disable heal and i/o is in
cb8e9e
progress on the fuse mount.
cb8e9e
cb8e9e
for more details refer following bug:
cb8e9e
https://bugzilla.redhat.com/show_bug.cgi?id=1259992
cb8e9e
https://bugzilla.redhat.com/show_bug.cgi?id=1260930
cb8e9e
cb8e9e
This is because of by default multi thread epoll in glusterd is 2.
cb8e9e
Workaround is to make epoll thread to 1 and set ping-timeout to 0
cb8e9e
cb8e9e
Change-Id: I34158b7f41c814f9bde85c7a32f4f0bc5dbaf86c
cb8e9e
BUG: 1262236
cb8e9e
Signed-off-by: Gaurav Kumar Garg <ggarg@redhat.com>
cb8e9e
Reviewed-on: https://code.engineering.redhat.com/gerrit/57629
cb8e9e
Reviewed-by: Krishnan Parthasarathi <kparthas@redhat.com>
cb8e9e
Tested-by: Krishnan Parthasarathi <kparthas@redhat.com>
cb8e9e
---
cb8e9e
 extras/glusterd.vol.in |    3 ++-
cb8e9e
 1 files changed, 2 insertions(+), 1 deletions(-)
cb8e9e
cb8e9e
diff --git a/extras/glusterd.vol.in b/extras/glusterd.vol.in
cb8e9e
index 690dbe7..316104c 100644
cb8e9e
--- a/extras/glusterd.vol.in
cb8e9e
+++ b/extras/glusterd.vol.in
cb8e9e
@@ -5,6 +5,7 @@ volume management
cb8e9e
     option transport.socket.keepalive-time 10
cb8e9e
     option transport.socket.keepalive-interval 2
cb8e9e
     option transport.socket.read-fail-log off
cb8e9e
-    option ping-timeout 30
cb8e9e
+    option ping-timeout 0
cb8e9e
+    option event-threads 1
cb8e9e
 #   option base-port 49152
cb8e9e
 end-volume
cb8e9e
-- 
cb8e9e
1.7.1
cb8e9e