readwritesplit.h:Removed invalid macros which assumed that ROUTER_CLIENT_SES->rses_backend_ref always pointed to valid and used backend reference and thus included potential risk of NULL-pointer refernce.
mysql_backend.c and mysql_client.c:avoid executing CHK_PROTOCOL(p) after original DCB has been released the memory.
readwritesplit.c:Replaced RSES_CLEINT_DCB macro with a function which returns client DCB for a given router client session.
tee.c:closeSession removed unnecessary dcb_free, router/service closes all backend DCBs and the client DCB, and client DCB is the one that was tried to free in closeSession.
readwritesplit.c:routeQuery now handles untyped and typed GWBUFs. Untyped means that read buffer may consist of incomplete and multiple MySQL packets. Typed buffer always consists of a single MySQL packet (which can be split to many buffers inside GWBUF).
Fixed Coverity cases #84840 and #84841
Execution of session commands failed because session commands and normal sql statements were executed in a wrong order if backend was executing previous session command while new sql stmt was routed to that backend. There was a window where ordering went wrong.
It is possible that one normal sql stmt arrives while previous sescmds are still being executed.
Introduced a new member in backend_ref_t structure, bref_pending_cmd where new sql stmt pointer is stored in that case. When sescmds are executed completely, that command is automatically executed next.
Added description of new rwsplit variables, read_ses_variables_from_slaves and write_ses_variables_to_all to MaxScale_template.cnf
readwritesplit.h:Set default values for write_ses_variables_to_all=true, and read_ses_variables_from_slaves=false
query_classifier.h: added QUERY_TYPE_CREATE_TMP_TABLE and QUERY_TYPE_READ_TMP_TABLE for use of temporary table support.
hashtable.c:Added variant of hashtable which is 'flat', that is, stored to existing memory instead of allocating memory as a part of the call. Existing function declarations don't change but added hashtable_alloc_flat for the purpose. Both hashtable_alloc and hashtable_alloc_flat now call the real allocation function, hashtable_alloc_real. hashtable_free only frees memory which is allocated in hashtable_alloc_real.
hashtable.h: added a flag to HASHTABLE struct to indicate whether hashtable owns its memory or not.
readwritesplit.h: Added RSES_PROP_TYPE_TMPTABLES property type to be used for keeping the hashtable for tablenames.
readwritesplit.c: Added comments about temporary table support implementation.
readwritesplit.h: Added TARGET_RLAG_MAX to the list of hint targets. It is used in case of hint max_slave_replication_lag=<value>
readwritesplit.c: get_dcb is called when readwritesplit router chooses target backend for a query. Backend is selected from those the router has connection to. get_dcb now takes additional parameters, name and max_rlag. Name is used if there is a hint route to server <server name>. max_rlag is used for hint max_slave_replication_lag=<value> or if it doesn't exist, then for configure parameter max_slave_replication_lag=<configured value> or if both are absent, ignored.
buffer.c: added memory release for hint of a GWBUF
hint.c: added bool hint_exists()
hint.h: added placeholder for hint type HINT_ROUTE_TO_ALL which doesn't have implementation yet.
filter/Makefile: fixed dependency issue
hintparser.c: added const char* token_get_keyword(), hint_parser:added NULL check, hint_next_token: fixed off-by-one bug
readwritesplit.h: added bitfield for hints' use, which includes route targets and flag for case where user hinted to route to named backend server.
readwritesplit.c: added function route_target_t get_route_target() for resolving route target based on 1) query type (from query_classifier) 2) transaction state (active/not) and 3) hints. Modified get_dcb, which is called in routeQuery to provide pointer to correct backend DCB. Now get_dcb also takes server unique name as a parameter if such a hint was found. for hints' use, which includes enter the commit message for your changes.
The flag is part of the active operation counting, which is utilized in load balancing. The active operation count per backend is used by default as criteria when router chooses to which backend a query should be routed.
All prepare commands are executed in every backend server currently connected.
All executes are routed to master. If stmt type was recorded in prepare phase in rwsplit router, read-only stmts could be routed to slaves.
COM_STMT_PREPARE gets arbitrary number of response packets from backend database. Since statements are prepared in every backend server and only one multi-packet response can be replied to client, redundant multi-packet responses are discarded. This is done in router. Mechanisms from session command handling are utilized with little changes: router must identify when response consists of multiple packets so that it knows to calculate the number of packets in response and that it is able to discard correct number of packets.
Information to the reply-handling router is provided by backend protocol, which includes a ordered list of commands of commands sent to protocol-owning backend server. A command is stored to protocol struct in mysql_backend.c:gw_MySQLWrite_backend if the statement buffer's type has GWBUF_TYPE_SINGLE_STMT set in mysql_client.c:route_by_statement. GWBUF_TYPE_SINGLE_STATEMENT indicates that there is single statement in the buffer, as opposite to Read Connection router, which accepts streaming input from client.
Added flags to those backend references which have sent something to backend which causes the backend to send results or reply back. Didn't add removal of the flag since there's currently no way to tell whether response from backend contains anything else than session command reply - which aren't counted when BREF_WAITING_RESULT is set and cleared.
Added a lot of logging to error, trace and message logs which should help the user to handle errors which can't be automatically resolved, like attempt to use nonexisting database.
monitors used mysql_ping without MYSQL_OPT_READ_TIMEOUT which caused read to block. Fixed in mysql and galera monitor.
Added log writing per each status change of each server and repeatedly if server is not running.
Removed SERVER_IS_JOINED checks from rwsplit router.
Added function pointer array which is indexed by enumerated selection criterias.
Selection criteria can now be switched by modifying config file and reloading it.
The option is reloadable in the same way as max_slave_connections parameter. Option value determines which sort function is called before to sort backend server pointers before selecting them and establishing connections to backends.
Added missing makefile to server/modules/routing/test . It is needed for system testing. Prior this fix, 'make testall' failed because the makefile was missing.
In practice, 'reload config' reads config file and updates config parameters of rwsplit router service. In addition to that there is a version number indicating what generation of configuration service holds. When router instance is created (when MaxScale is started) service's config version is copied to router intance. Whenever new client connection (rwsplit session) is started, router instance's config version is checked against that of service's. If versions differ, service's config data is copied to router instance. New session will be started with router instance's config values.
1. in query_classifier.cc autocommit_enabled, and transaction_active variables maintained their values across different sessions. Now those values are stored in each router_client_ses object.
2. As a part of implementation of MAX-95 session variables were added to BACKEND struct which is shared with all sessions using the SERVICE which the particular BACKEND serves. Now each router_client_ses object has a backend reference struct which includes pointer to BACKEND, DCB and to session command cursor.
Added test - set_autocommit_disabled.sql, test_after_autocommit_disabled.sql - to check that session variable is discarded when session where it belongs terminates.
Parameter it read from config file to CONFIG_CONTEXT's parameter list. It is qualified in service.c:service_set_slave_conn_limit and if qualified, the qualified integer value and the value type are copied to the CONFIG_PARAMETER structure.
This CONFIG_PARAMETER struct is cloned (=copied to different memory area) and linked to RW Split SERVICE struct.
When RW Split router_instance is created in readwritesplit.c:createInstance, the value is copied to (new) rwsplit_config_t structure from SERVICE's parameter list.
When new routing session is created in readwritesplit.c:newSession, the rwsplit_config_t structure is copied to ROUTER_CLIENT_SES struct and the actual max_nslaves value is calculated from the config value (if percentage is used).
Tests and many error handling branches are missing but functionality seems to be working.
This implementation doesn't guarantee execution order between session commands and queries
if other backend server lags behind in session command execution.
In poll.c : moved processing of EPOLLERR and EPOLLHUP after processing of EPOLLIN and EPOLLOUT.
This ensures that COM_QUIT messages are read and routed forward before signals arrive (from local client/backend).
nection router stream is used and with read/write split router individual statements are passed to router.
Added new function to ROUTER_OBJECT : uint8_t (*getCapabilities)(ROUTER *instance, void* router_session); which is implemented in every route
r.
Added support for multi-statement packets in rwsplit router. In other words, if network packet includes multiple mysql statements, they are separated and passed to router one by one.
Multi-packet statements (those which exceeds network packet boundaries) are _not_ supported yet.
