asterisk/res/res_sorcery_astdb.c

435 lines
14 KiB
C
Raw Permalink Normal View History

/*
* Asterisk -- An open source telephony toolkit.
*
* Copyright (C) 2013, Digium, Inc.
*
* Joshua Colp <jcolp@digium.com>
*
* See http://www.asterisk.org for more information about
* the Asterisk project. Please do not directly contact
* any of the maintainers of this project for assistance;
* the project provides a web site, mailing lists and IRC
* channels for your use.
*
* This program is free software, distributed under the terms of
* the GNU General Public License Version 2. See the LICENSE file
* at the top of the source tree.
*/
/*!
* \file
*
* \brief Sorcery Astdb Object Wizard
*
* \author Joshua Colp <jcolp@digium.com>
*/
/*** MODULEINFO
<support_level>core</support_level>
***/
#include "asterisk.h"
#include <regex.h>
#include "asterisk/module.h"
#include "asterisk/sorcery.h"
#include "asterisk/astdb.h"
#include "asterisk/json.h"
static void *sorcery_astdb_open(const char *data);
static int sorcery_astdb_create(const struct ast_sorcery *sorcery, void *data, void *object);
static void *sorcery_astdb_retrieve_id(const struct ast_sorcery *sorcery, void *data, const char *type, const char *id);
static void *sorcery_astdb_retrieve_fields(const struct ast_sorcery *sorcery, void *data, const char *type, const struct ast_variable *fields);
static void sorcery_astdb_retrieve_multiple(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects,
const struct ast_variable *fields);
static void sorcery_astdb_retrieve_regex(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects, const char *regex);
static void sorcery_astdb_retrieve_prefix(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects, const char *prefix, const size_t prefix_len);
static int sorcery_astdb_update(const struct ast_sorcery *sorcery, void *data, void *object);
static int sorcery_astdb_delete(const struct ast_sorcery *sorcery, void *data, void *object);
static void sorcery_astdb_close(void *data);
static struct ast_sorcery_wizard astdb_object_wizard = {
.name = "astdb",
.open = sorcery_astdb_open,
.create = sorcery_astdb_create,
.retrieve_id = sorcery_astdb_retrieve_id,
.retrieve_fields = sorcery_astdb_retrieve_fields,
.retrieve_multiple = sorcery_astdb_retrieve_multiple,
.retrieve_regex = sorcery_astdb_retrieve_regex,
.retrieve_prefix = sorcery_astdb_retrieve_prefix,
.update = sorcery_astdb_update,
.delete = sorcery_astdb_delete,
.close = sorcery_astdb_close,
};
static int sorcery_astdb_create(const struct ast_sorcery *sorcery, void *data, void *object)
{
RAII_VAR(struct ast_json *, objset, ast_sorcery_objectset_json_create(sorcery, object), ast_json_unref);
RAII_VAR(char *, value, NULL, ast_json_free);
const char *prefix = data;
char family[strlen(prefix) + strlen(ast_sorcery_object_get_type(object)) + 2];
if (!objset || !(value = ast_json_dump_string(objset))) {
return -1;
}
snprintf(family, sizeof(family), "%s/%s", prefix, ast_sorcery_object_get_type(object));
return ast_db_put(family, ast_sorcery_object_get_id(object), value);
}
/*! \brief Internal helper function which returns a filtered objectset.
*
* The following are filtered out of the objectset:
* \li Fields that are not registered with sorcery.
*
* \param objectset Objectset to filter.
* \param sorcery The sorcery instance that is requesting an objectset.
* \param type The object type
*
* \return The filtered objectset
*/
static struct ast_variable *sorcery_astdb_filter_objectset(struct ast_variable *objectset, const struct ast_sorcery *sorcery,
const char *type)
{
struct ast_variable *previous = NULL, *field = objectset;
struct ast_sorcery_object_type *object_type;
object_type = ast_sorcery_get_object_type(sorcery, type);
if (!object_type) {
ast_log(LOG_WARNING, "Unknown sorcery object type %s. Expect errors\n", type);
return objectset;
}
while (field) {
struct ast_variable *removed;
if (ast_sorcery_is_object_field_registered(object_type, field->name)) {
previous = field;
field = field->next;
continue;
}
ast_debug(1, "Filtering out astdb field '%s' from retrieval\n", field->name);
if (previous) {
previous->next = field->next;
} else {
objectset = field->next;
}
removed = field;
field = field->next;
removed->next = NULL;
ast_variables_destroy(removed);
}
ao2_cleanup(object_type);
return objectset;
}
/*! \brief Internal helper function which retrieves an object, or multiple objects, using fields for criteria */
static void *sorcery_astdb_retrieve_fields_common(const struct ast_sorcery *sorcery, void *data, const char *type, const struct ast_variable *fields, struct ao2_container *objects)
{
const char *prefix = data;
char family[strlen(prefix) + strlen(type) + 2];
RAII_VAR(struct ast_db_entry *, entries, NULL, ast_db_freetree);
struct ast_db_entry *entry;
snprintf(family, sizeof(family), "%s/%s", prefix, type);
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
if (!(entries = ast_db_gettree(family, NULL))) {
return NULL;
}
for (entry = entries; entry; entry = entry->next) {
const char *key = entry->key + strlen(family) + 2;
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
struct ast_json_error error;
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
RAII_VAR(struct ast_variable *, existing, NULL, ast_variables_destroy);
void *object = NULL;
if (!(json = ast_json_load_string(entry->data, &error))) {
return NULL;
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
}
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
if (ast_json_to_ast_variables(json, &existing) != AST_JSON_TO_AST_VARS_CODE_SUCCESS) {
return NULL;
}
existing = sorcery_astdb_filter_objectset(existing, sorcery, type);
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
if (fields && !ast_variable_lists_match(existing, fields, 0)) {
continue;
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
}
if (!(object = ast_sorcery_alloc(sorcery, type, key)) ||
ast_sorcery_objectset_apply(sorcery, object, existing)) {
ao2_cleanup(object);
return NULL;
}
if (!objects) {
return object;
}
ao2_link(objects, object);
ao2_cleanup(object);
}
return NULL;
}
static void *sorcery_astdb_retrieve_fields(const struct ast_sorcery *sorcery, void *data, const char *type, const struct ast_variable *fields)
{
return sorcery_astdb_retrieve_fields_common(sorcery, data, type, fields, NULL);
}
static void *sorcery_astdb_retrieve_id(const struct ast_sorcery *sorcery, void *data, const char *type, const char *id)
{
const char *prefix = data;
char family[strlen(prefix) + strlen(type) + 2];
RAII_VAR(char *, value, NULL, ast_free_ptr);
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
struct ast_json_error error;
RAII_VAR(struct ast_variable *, objset, NULL, ast_variables_destroy);
void *object = NULL;
snprintf(family, sizeof(family), "%s/%s", prefix, type);
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
if (ast_db_get_allocated(family, id, &value)
|| !(json = ast_json_load_string(value, &error))
|| (ast_json_to_ast_variables(json, &objset) != AST_JSON_TO_AST_VARS_CODE_SUCCESS)
|| !(objset = sorcery_astdb_filter_objectset(objset, sorcery, type))
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
|| !(object = ast_sorcery_alloc(sorcery, type, id))
|| ast_sorcery_objectset_apply(sorcery, object, objset)) {
ast_debug(3, "Failed to retrieve object '%s' from astdb\n", id);
ao2_cleanup(object);
return NULL;
}
return object;
}
static void sorcery_astdb_retrieve_multiple(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects, const struct ast_variable *fields)
{
sorcery_astdb_retrieve_fields_common(sorcery, data, type, fields, objects);
}
/*!
* \internal
* \brief Convert regex prefix pattern to astDB prefix pattern if possible.
*
* \param tree astDB prefix pattern buffer to fill.
* \param regex Extended regular expression with the start anchor character '^'.
*
* \note Since this is a helper function, the tree buffer is
* assumed to always be large enough.
*
* \retval 0 on success.
* \retval -1 on error. regex is invalid.
*/
static int make_astdb_prefix_pattern(char *tree, const char *regex)
{
const char *src;
char *dst;
for (dst = tree, src = regex + 1; *src; ++src) {
if (*src == '\\') {
/* Escaped regex char. */
++src;
if (!*src) {
/* Invalid regex. The caller escaped the string terminator. */
return -1;
}
} else if (*src == '$') {
if (!src[1]) {
/* Remove the tail anchor character. */
*dst = '\0';
return 0;
}
} else if (strchr(".?*+{[(|", *src)) {
/*
* The regex is not a simple prefix pattern.
*
* XXX With more logic, it is possible to simply
* use the current prefix pattern. The last character
* needs to be removed if possible when the current regex
* token is "?*{". Also the rest of the regex pattern
* would need to be checked for subgroup/alternation.
* Subgroup/alternation is too complex for a simple prefix
* match.
*/
dst = tree;
break;
}
*dst++ = *src;
}
if (dst != tree) {
*dst++ = '%';
}
*dst = '\0';
return 0;
}
static void sorcery_astdb_retrieve_regex(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects, const char *regex)
{
const char *prefix = data;
char family[strlen(prefix) + strlen(type) + 2];
char tree[strlen(regex) + 1];
RAII_VAR(struct ast_db_entry *, entries, NULL, ast_db_freetree);
regex_t expression;
struct ast_db_entry *entry;
snprintf(family, sizeof(family), "%s/%s", prefix, type);
if (regex[0] == '^') {
/*
* For performance reasons, try to create an astDB prefix
* pattern from the regex to reduce the number of entries
* retrieved from astDB for regex to then match.
*/
if (make_astdb_prefix_pattern(tree, regex)) {
return;
}
} else {
tree[0] = '\0';
}
if (!(entries = ast_db_gettree(family, tree))
|| regcomp(&expression, regex, REG_EXTENDED | REG_NOSUB)) {
return;
}
for (entry = entries; entry; entry = entry->next) {
/* The key in the entry includes the family, so we need to strip it out for regex purposes */
const char *key = entry->key + strlen(family) + 2;
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
struct ast_json_error error;
RAII_VAR(void *, object, NULL, ao2_cleanup);
RAII_VAR(struct ast_variable *, objset, NULL, ast_variables_destroy);
if (regexec(&expression, key, 0, NULL, 0)) {
continue;
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
} else if (!(json = ast_json_load_string(entry->data, &error))
|| (ast_json_to_ast_variables(json, &objset) != AST_JSON_TO_AST_VARS_CODE_SUCCESS)
|| !(objset = sorcery_astdb_filter_objectset(objset, sorcery, type))
sorcery/res_pjsip: Refactor for realtime performance There were a number of places in the res_pjsip stack that were getting all endpoints or all aors, and then filtering them locally. A good example is pjsip_options which, on startup, retrieves all endpoints, then the aors for those endpoints, then tests the aors to see if the qualify_frequency is > 0. One issue was that it never did anything with the endpoints other than retrieve the aors so we probably could have skipped a step and just retrieved all aors. But nevermind. This worked reasonably well with local config files but with a realtime backend and thousands of objects, this was a nightmare. The issue really boiled down to the fact that while realtime supports predicates that are passed to the database engine, the non-realtime sorcery backends didn't. They do now. The realtime engines have a scheme for doing simple comparisons. They take in an ast_variable (or list) for matching, and the name of each variable can contain an operator. For instance, a name of "qualify_frequency >" and a value of "0" would create a SQL predicate that looks like "where qualify_frequency > '0'". If there's no operator after the name, the engines add an '=' so a simple name of "qualify_frequency" and a value of "10" would return exact matches. The non-realtime backends decide whether to include an object in a result set by calling ast_sorcery_changeset_create on every object in the internal container. However, ast_sorcery_changeset_create only does exact string matches though so a name of "qualify_frequency >" and a value of "0" returns nothing because the literal "qualify_frequency >" doesn't match any name in the objset set. So, the real task was to create a generic string matcher that can take a left value, operator and a right value and perform the match. To that end, strings.c has a new ast_strings_match(left, operator, right) function. Left and right are the strings to operate on and the operator can be a string containing any of the following: = (or NULL or ""), !=, >, >=, <, <=, like or regex. If the operator is like or regex, the right string should be a %-pattern or a regex expression. If both left and right can be converted to float, then a numeric comparison is performed, otherwise a string comparison is performed. To use this new function on ast_variables, 2 new functions were added to config.c. One that compares 2 ast_variables, and one that compares 2 ast_variable lists. The former is useful when you want to compare 2 ast_variables that happen to be in a list but don't want to traverse the list. The latter will traverse the right list and return true if all the variables in it match the left list. Now, the backends' fields_cmp functions call ast_variable_lists_match instead of ast_sorcery_changeset_create and they can now process the same syntax as the realtime engines. The realtime backend just passes the variable list unaltered to the engine. The only gotcha is that there's no common realtime engine support for regex so that's been noted in the api docs for ast_sorcery_retrieve_by_fields. Only one more change to sorcery was done... A new config flag "allow_unqualified_fetch" was added to reg_sorcery_realtime. "no": ignore fetches if no predicate fields were supplied. "error": same as no but emit an error. (good for testing) "yes": allow (the default); "warn": allow but emit a warning. (good for testing) Now on to res_pjsip... pjsip_options was modified to retrieve aors with qualify_frequency > 0 rather than all endpoints then all aors. Not only was this a big improvement in realtime retrieval but even for config files there's an improvement because we're not going through endpoints anymore. res_pjsip_mwi was modified to retieve only endpoints with something in the mailboxes field instead of all endpoints then testing mailboxes. res_pjsip_registrar_expire was completely refactored. It was retrieving all contacts then setting up scheduler entries to check for expiration. Now, it's a single thread (like keepalive) that periodically retrieves only contacts whose expiration time is < now and deletes them. A new contact_expiration_check_interval was added to global with a default of 30 seconds. Ross Beer reports that with this patch, his Asterisk startup time dropped from around an hour to under 30 seconds. There are still objects that can't be filtered at the database like identifies, transports, and registrations. These are not going to be anywhere near as numerous as endpoints, aors, auths, contacts however. Back to allow_unqualified_fetch. If this is set to yes and you have a very large number of objects in the database, the pjsip CLI commands will attempt to retrive ALL of them if not qualified with a LIKE. Worse, if you type "pjsip show endpoint <tab>" guess what's going to happen? :) Having a cache helps but all the objects will have to be retrieved at least once to fill the cache. Setting allow_unqualified_fetch=no prevents the mass retrieve and should be used on endpoints, auths, aors, and contacts. It should NOT be used for identifies, registrations and transports since these MUST be retrieved in bulk. Example sorcery.conf: [res_pjsip] endpoint=config,pjsip.conf,criteria=type=endpoint endpoint=realtime,ps_endpoints,allow_unqualified_fetch=error ASTERISK-25826 #close Reported-by: Ross Beer Tested-by: Ross Beer Change-Id: Id2691e447db90892890036e663aaf907b2dc1c67
2016-03-08 21:55:30 +00:00
|| !(object = ast_sorcery_alloc(sorcery, type, key))
|| ast_sorcery_objectset_apply(sorcery, object, objset)) {
regfree(&expression);
return;
}
ao2_link(objects, object);
}
regfree(&expression);
}
static void sorcery_astdb_retrieve_prefix(const struct ast_sorcery *sorcery, void *data, const char *type, struct ao2_container *objects, const char *prefix, const size_t prefix_len)
{
const char *family_prefix = data;
size_t family_len = strlen(family_prefix) + strlen(type) + 1; /* +1 for slash delimiter */
char family[family_len + 1];
char tree[prefix_len + 1];
RAII_VAR(struct ast_db_entry *, entries, NULL, ast_db_freetree);
struct ast_db_entry *entry;
snprintf(tree, sizeof(tree), "%.*s", (int) prefix_len, prefix);
snprintf(family, sizeof(family), "%s/%s", family_prefix, type);
if (!(entries = ast_db_gettree_by_prefix(family, tree))) {
return;
}
for (entry = entries; entry; entry = entry->next) {
/* The key in the entry includes the family, so we need to strip it out */
const char *key = entry->key + family_len + 2;
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
struct ast_json_error error;
RAII_VAR(void *, object, NULL, ao2_cleanup);
RAII_VAR(struct ast_variable *, objset, NULL, ast_variables_destroy);
if (!(json = ast_json_load_string(entry->data, &error))
|| (ast_json_to_ast_variables(json, &objset) != AST_JSON_TO_AST_VARS_CODE_SUCCESS)
|| !(objset = sorcery_astdb_filter_objectset(objset, sorcery, type))
|| !(object = ast_sorcery_alloc(sorcery, type, key))
|| ast_sorcery_objectset_apply(sorcery, object, objset)) {
return;
}
ao2_link(objects, object);
}
}
static int sorcery_astdb_update(const struct ast_sorcery *sorcery, void *data, void *object)
{
const char *prefix = data;
char family[strlen(prefix) + strlen(ast_sorcery_object_get_type(object)) + 2], value[2];
snprintf(family, sizeof(family), "%s/%s", prefix, ast_sorcery_object_get_type(object));
/* It is okay for the value to be truncated, we are only checking that it exists */
if (ast_db_get(family, ast_sorcery_object_get_id(object), value, sizeof(value))) {
return -1;
}
/* The only difference between update and create is that for update the object must already exist */
return sorcery_astdb_create(sorcery, data, object);
}
static int sorcery_astdb_delete(const struct ast_sorcery *sorcery, void *data, void *object)
{
const char *prefix = data;
char family[strlen(prefix) + strlen(ast_sorcery_object_get_type(object)) + 2];
char value[2];
snprintf(family, sizeof(family), "%s/%s", prefix, ast_sorcery_object_get_type(object));
if (ast_db_get(family, ast_sorcery_object_get_id(object), value, sizeof(value))) {
return -1;
}
return ast_db_del(family, ast_sorcery_object_get_id(object));
}
static void *sorcery_astdb_open(const char *data)
{
/* We require a prefix for family string generation, or else stuff could mix together */
if (ast_strlen_zero(data)) {
return NULL;
}
return ast_strdup(data);
}
static void sorcery_astdb_close(void *data)
{
ast_free(data);
}
static int load_module(void)
{
if (ast_sorcery_wizard_register(&astdb_object_wizard)) {
return AST_MODULE_LOAD_DECLINE;
}
return AST_MODULE_LOAD_SUCCESS;
}
static int unload_module(void)
{
ast_sorcery_wizard_unregister(&astdb_object_wizard);
return 0;
}
AST_MODULE_INFO(ASTERISK_GPL_KEY, AST_MODFLAG_GLOBAL_SYMBOLS | AST_MODFLAG_LOAD_ORDER, "Sorcery Astdb Object Wizard",
.support_level = AST_MODULE_SUPPORT_CORE,
.load = load_module,
.unload = unload_module,
.load_pri = AST_MODPRI_REALTIME_DRIVER,
);