[Gluster-devel] all-null pending matrix again

[Emmanuel Dreyfus]

On Mon, Sep 30, 2013 at 07:28:36AM +0000, Emmanuel Dreyfus wrote:
> Right... In the meantime I guess need to completely disable eager-locks 
> oon netBSD package.

I tried this without success. I am missing somethin?

--- xlators/cluster/afr/src/afr.c.orig  2013-09-30 09:33:14.000000000 +0200
+++ xlators/cluster/afr/src/afr.c       2013-09-30 09:35:55.000000000 +0200
@@ -175,8 +175,21 @@
                 priv->read_child = index;
         }
 
         GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
+
+#ifdef __NetBSD__
+       /*
+        * eager-locks are broken on NetBSD, and cause spurious
+        * splut brain with all NULL pending matrix, as described
+        * in https://bugzilla.redhat.com/show_bug.cgi?id=1005526
+        * We therefore disable them for now
+        */
+       if (priv->eager_lock != 0) {
+               gf_log (this->name, GF_LOG_WARNING, "eager locks disabled");
+               priv->eager_lock = 0;
+       }
+#endif
         GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
         GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
                           uint32, out);
         fix_quorum_options(this,priv,qtype);


-- 
Emmanuel Dreyfus
manu at netbsd.org