Re: (ITS#7222) Slapd hangs on high load

[hrvoje.habjanic@zg.t-com.hr]

This is a multi-part message in MIME format.
--------------040402080004020608060006
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

On 09.04.2012 20:09, Hrvoje HabjaniÄ? wrote:
>
> Hi.
>
> Two more "hang"s, both in sched_yield(). This is with replacement malloc
> (tcmalloc, minimal).
>
> http://free-zg.t-com.hr/HrvojeHabjanic/openldap/hang4.log
> http://free-zg.t-com.hr/HrvojeHabjanic/openldap/hang5.log
>
> H.

Hi.

Attached patch if solving my proglem with "hang" in sched_yield.

In general, i do think that there (cache management) is a lot of
unnecessary locking and waiting ... And simplifying things there would
solve a lot of problems ... Probably. :-)

Of course, i'm not shure how will this change influence the rest of the
code, but it does work for me (tm).

H.

p.s. Also available at
http://free-zg.t-com.hr/HrvojeHabjanic/openldap/ol.diff


--------------040402080004020608060006
Content-Type: text/x-patch;
 name="ol.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="ol.diff"

diff -Nru openldap-d7d1789-org/servers/slapd/back-bdb/cache.c openldap-d7d1789/servers/slapd/back-bdb/cache.c
--- openldap-d7d1789-org/servers/slapd/back-bdb/cache.c	2012-04-10 00:47:23.000000000 +0200
+++ openldap-d7d1789/servers/slapd/back-bdb/cache.c	2012-04-11 10:23:43.267461183 +0200
@@ -1361,7 +1361,7 @@
     DB_LOCK	*lock )
 {
 	EntryInfo *ei = BEI(e);
-	int	rc, busy = 0;
+	int	rc, busy = 0, counter = 0;
 
 	assert( e->e_private != NULL );
 
@@ -1378,7 +1378,7 @@
 
 	bdb_cache_entryinfo_unlock( ei );
 
-	while ( busy ) {
+	while ( busy && counter < 1000) {
 		ldap_pvt_thread_yield();
 		busy = 0;
 		bdb_cache_entryinfo_lock( ei );
@@ -1387,6 +1387,13 @@
 			ei->bei_finders > 0 )
 			busy = 1;
 		bdb_cache_entryinfo_unlock( ei );
+		counter ++;
+	}
+	if( busy ) {
+		bdb_cache_entryinfo_lock( ei );
+		ei->bei_state ^= CACHE_ENTRY_DELETED;
+		bdb_cache_entryinfo_unlock( ei );
+		return DB_LOCK_DEADLOCK;
 	}
 
 	/* Get write lock on the data */

--------------040402080004020608060006--