db_cluster: new parameter max_query_length

- set threshold in second after which a connection is marked inactive if
  the query fails
- allow to let connectioncs being active if the error is due to some sql
  mistake
- default is 0 (set connection innactive immediately, same behavior as
  so far)

modules/db_cluster/README

@@ -28,6 +28,7 @@ Daniel-Constantin Mierla
               3.1. connection (str)
               3.2. cluster (str)
               3.3. inactive_interval (int)
+              3.4. max_query_length (int)
 
         4. Usage
 
@@ -36,7 +37,8 @@ Daniel-Constantin Mierla
    1.1. Set connection parameter
    1.2. Set cluster parameter
    1.3. Set inactive_interval parameter
-   1.4. Set cluster parameter
+   1.4. Set max_query_length parameter
+   1.5. Set cluster parameter
 
 Chapter 1. Admin Guide
 
@@ -53,6 +55,7 @@ Chapter 1. Admin Guide
         3.1. connection (str)
         3.2. cluster (str)
         3.3. inactive_interval (int)
+        3.4. max_query_length (int)
 
    4. Usage
 
@@ -90,6 +93,7 @@ Chapter 1. Admin Guide
    3.1. connection (str)
    3.2. cluster (str)
    3.3. inactive_interval (int)
+   3.4. max_query_length (int)
 
 3.1. connection (str)
 
@@ -148,6 +152,21 @@ modparam("db_cluster", "cluster", "cls1=>con1=9s8p;con2=9s8p")
 modparam("db_cluster", "inactive_interval", 180)
 ...
 
+3.4. max_query_length (int)
+
+   How long (seconds) a failed db operation needs to last before
+   deactivating the connection for inactive_interval seconds. This
+   prevents disabling of connections that reply fast with error codes,
+   thus being active (e.g., due to primary key insert errors). In such
+   cases, the database server is active.
+
+   Default value is 0.
+
+   Example 1.4. Set max_query_length parameter
+...
+modparam("db_cluster", "max_query_length", 5)
+...
+
 4. Usage
 
    Practically, all the modules that want to use a cluster, have to set
@@ -169,7 +188,7 @@ modparam("db_cluster", "inactive_interval", 180)
    to be used for parallel writing from acc and round-robin reading by
    sqlops.
 
-   Example 1.4. Set cluster parameter
+   Example 1.5. Set cluster parameter
 ...
 modparam("db_cluster", "connection",
              "c1=>mysql://kamailio:kamailiorw@localhost/kamailio1")

modules/db_cluster/db_cluster_mod.c

@@ -37,6 +37,7 @@ int dbcl_con_param(modparam_t type, void *val);
 int dbcl_cls_param(modparam_t type, void *val);
 
 int dbcl_inactive_interval = 300;
+int dbcl_max_query_length = 0;
 
 /*! \brief
  * DB Cluster module interface
@@ -53,6 +54,7 @@ static param_export_t params[] = {
 	{"connection",  STR_PARAM|USE_FUNC_PARAM, (void*)dbcl_con_param},
 	{"cluster",     STR_PARAM|USE_FUNC_PARAM, (void*)dbcl_cls_param},
 	{"inactive_interval",     INT_PARAM,    &dbcl_inactive_interval},
+	{"max_query_length",     INT_PARAM,    &dbcl_max_query_length},
 	{0, 0, 0}
 };
 

modules/db_cluster/dbcl_api.c

@@ -33,10 +33,12 @@
 #include "../../trim.h"
 #include "../../globals.h"
 #include "../../lib/srdb1/db.h"
+#include "../../timer.h"
 
 #include "dbcl_data.h"
 #include "dbcl_api.h"
 
+extern int dbcl_max_query_length;
 
 #define DBCL_READ(command) \
 	do {\
@@ -44,6 +46,7 @@
 	int i;\
 	int j;\
 	int k;\
+	unsigned int sec = 0;\
 	db1_con_t  *dbh=NULL;\
 	dbcl_cls_t *cls=NULL;\
 	cls = (dbcl_cls_t*)_h->tail;\
@@ -60,6 +63,7 @@
 					{\
 						LM_DBG("serial operation - cluster [%.*s] (%d/%d)\n",\
 								cls->name.len, cls->name.s, i, j);\
+						sec = get_ticks();\
 						dbh = cls->rlist[i].clist[j]->dbh;\
 						ret = cls->rlist[i].clist[j]->dbf.command;\
 						if (ret==0) {\
@@ -69,7 +73,10 @@
 							LM_DBG("serial operation - failre on cluster"\
 									" [%.*s] (%d/%d)\n",\
 									cls->name.len, cls->name.s, i, j);\
-							dbcl_inactive_con(cls->rlist[i].clist[j]);\
+							sec = get_ticks() - sec;\
+							if(sec >= dbcl_max_query_length){\
+								dbcl_inactive_con(cls->rlist[i].clist[j]);\
+							}\
 						}\
 					}\
 				}\
@@ -83,6 +90,7 @@
 					{\
 						LM_DBG("round robin operation - cluster [%.*s] (%d/%d)\n",\
 								cls->name.len, cls->name.s, i, j);\
+						sec = get_ticks();\
 						dbh = cls->rlist[i].clist[j]->dbh;\
 						ret = cls->rlist[i].clist[j]->dbf.command;\
 						if (ret==0)\
@@ -94,7 +102,10 @@
 							LM_DBG("round robin operation - failre on cluster"\
 									" [%.*s] (%d/%d)\n",\
 									cls->name.len, cls->name.s, i, j);\
-							dbcl_inactive_con(cls->rlist[i].clist[j]);\
+							sec = get_ticks() - sec;\
+							if(sec >= dbcl_max_query_length){\
+								dbcl_inactive_con(cls->rlist[i].clist[j]);\
+							}\
 						}\
 					}\
 				}\
@@ -118,6 +129,7 @@
 	int i;\
 	int j;\
 	int k;\
+	unsigned int sec = 0;\
 	db1_con_t  *dbh=NULL;\
 	dbcl_cls_t *cls=NULL;\
 	cls = (dbcl_cls_t*)_h->tail;\
@@ -136,6 +148,7 @@
 					{\
 						LM_DBG("serial operation - cluster [%.*s] (%d/%d)\n",\
 								cls->name.len, cls->name.s, i, j);\
+						sec = get_ticks();\
 						dbh = cls->wlist[i].clist[j]->dbh;\
 						ret = cls->wlist[i].clist[j]->dbf.command;\
 						if (ret==0) {\
@@ -145,7 +158,10 @@
 							LM_DBG("serial operation - failure on cluster"\
 									" [%.*s] (%d/%d)\n",\
 									cls->name.len, cls->name.s, i, j);\
-							dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							sec = get_ticks() - sec;\
+							if(sec >= dbcl_max_query_length){\
+								dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							}\
 						}\
 					}\
 				}\
@@ -159,6 +175,7 @@
 					{\
 						LM_DBG("round robin operation - cluster [%.*s] (%d/%d)\n",\
 								cls->name.len, cls->name.s, i, j);\
+						sec = get_ticks();\
 						dbh = cls->wlist[i].clist[j]->dbh;\
 						ret = cls->wlist[i].clist[j]->dbf.command;\
 						if (ret==0)\
@@ -170,7 +187,10 @@
 							LM_DBG("round robin operation - failure on cluster"\
 									" [%.*s] (%d/%d)\n",\
 									cls->name.len, cls->name.s, i, j);\
-							dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							sec = get_ticks() - sec;\
+							if(sec >= dbcl_max_query_length){\
+								dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							}\
 						}\
 					}\
 				}\
@@ -183,6 +203,7 @@
 					{\
 						LM_DBG("parallel operation - cluster [%.*s] (%d/%d)\n",\
 								cls->name.len, cls->name.s, i, j);\
+						sec = get_ticks();\
 						dbh = cls->wlist[i].clist[j]->dbh;\
 						rc = cls->wlist[i].clist[j]->dbf.command;\
 						if(rc==0) {\
@@ -192,7 +213,10 @@
 							LM_DBG("parallel operation - failure on cluster"\
 									" [%.*s] (%d/%d)\n",\
 									cls->name.len, cls->name.s, i, j);\
-							dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							sec = get_ticks() - sec;\
+							if(sec >= dbcl_max_query_length){\
+								dbcl_inactive_con(cls->wlist[i].clist[j]);\
+							}\
 						}\
 						ret |= rc;\
 					}\

modules/db_cluster/doc/db_cluster_admin.xml

@@ -143,6 +143,29 @@ modparam("db_cluster", "cluster", "cls1=>con1=9s8p;con2=9s8p")
 ...
 modparam("db_cluster", "inactive_interval", 180)
 ...
+</programlisting>
+		</example>
+	</section>
+	<section id="db_cluster.p.max_query_length">
+		<title><varname>max_query_length</varname> (int)</title>
+		<para>
+		How long (seconds) a failed db operation needs to last before
+		deactivating the connection for inactive_interval seconds. This
+		prevents disabling of connections that reply fast with error
+		codes, thus being active (e.g., due to primary key insert errors).
+		In such cases, the database server is active.
+		</para>
+		<para>
+		<emphasis>
+			Default value is 0.
+		</emphasis>
+		</para>
+		<example>
+		<title>Set <varname>max_query_length</varname> parameter</title>
+		<programlisting format="linespecific">
+...
+modparam("db_cluster", "max_query_length", 5)
+...
 </programlisting>
 		</example>
 	</section>