hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix binlogrouter deadlock on reconnect

Browse Source 
The master reconnection in the binlogrouter used the housekeeper
recursively. This is no longer supported and the delayed worker call
mechanism should be used in its stead.
This commit is contained in:
Markus Mäkelä
2018-08-29 11:03:51 +03:00
parent 54dca0963a
commit 178d31af27
3 changed files with 21 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files

81
server/modules/routing/binlogrouter/blr_master.cc
View File

@ -47,6 +47,7 @@
#include <maxscale/protocol/mysql.h>
#include <maxscale/router.h>
#include <maxscale/routingworker.h>
#include <maxbase/worker.hh>
#include <maxscale/server.h>
#include <maxscale/service.h>
#include <maxscale/session.h>
@ -110,7 +111,6 @@ extern int blr_write_special_event(ROUTER_INSTANCE *router,
extern int blr_file_new_binlog(ROUTER_INSTANCE *router, char *file);
static bool blr_handle_missing_files(ROUTER_INSTANCE *router,
char *new_file);
static void worker_cb_start_master(MXB_WORKER*, void* data);
extern void blr_file_update_gtid(ROUTER_INSTANCE *router);
static int blr_check_connect_retry(ROUTER_INSTANCE *router);
@ -222,13 +222,8 @@ static void blr_start_master(void* data)
spinlock_acquire(&router->lock);
router->retry_count++;
spinlock_release(&router->lock);
/* Set reconnection task */
std::string name = router->service->name;
name += " Master";
hktask_add(name.c_str(),
blr_start_master_in_main,
router,
connect_retry);
blr_start_master_in_main(router, connect_retry);
MXS_ERROR("%s: failure while connecting to master server '%s', "
"retrying in %d seconds",
@ -283,12 +278,12 @@ static void blr_start_master(void* data)
* @param worker The worker in whose context the function is called.
* @param data The data to be passed to `blr_start_master`
*/
static void worker_cb_start_master(MXB_WORKER* worker, void* data)
static bool worker_cb_start_master(mxb::Worker::Call::action_t, ROUTER_INSTANCE* data)
{
// This is itended to be called only in the main worker.
mxb_assert(worker == mxs_rworker_get(MXS_RWORKER_MAIN));
mxb_assert_message(mxs_rworker_get_current() == mxs_rworker_get(MXS_RWORKER_MAIN),
"worker_cb_start_master must be called from the main thread");
blr_start_master(data);
return false;
}
/**
@ -296,21 +291,15 @@ static void worker_cb_start_master(MXB_WORKER* worker, void* data)
*
* @param data Data intended for `blr_start_master`.
*/
bool blr_start_master_in_main(void* data)
bool blr_start_master_in_main(ROUTER_INSTANCE* data, int32_t delay)
{
// The master should be connected to in the main worker, so we post it a
// message and call `blr_start_master` there.
MXB_WORKER* worker = mxs_rworker_get(MXS_RWORKER_MAIN); // The worker running in the main thread.
mxb::Worker* worker = (mxb::Worker*)mxs_rworker_get(MXS_RWORKER_MAIN);
mxb_assert(worker);
intptr_t arg1 = (intptr_t)worker_cb_start_master;
intptr_t arg2 = (intptr_t)data;
if (!mxb_worker_post_message(worker, MXB_WORKER_MSG_CALL, arg1, arg2))
{
MXS_ERROR("Could not post 'blr_start_master' message to main worker.");
}
worker->delayed_call(delay == 0 ? 1 : delay * 1000, worker_cb_start_master, data);
return false;
}
@ -390,26 +379,13 @@ blr_restart_master(ROUTER_INSTANCE *router)
router->retry_count++;
spinlock_release(&router->lock);
/* Set reconnection task */
static const char master[] = "Master";
size_t sz = strlen(router->service->name) + sizeof(master) + 2;
char *name = (char *)MXS_MALLOC(sz);
blr_start_master_in_main(router, connect_retry);
if (name)
{
snprintf(name, sz, "%s %s", router->service->name, master);
hktask_add(name,
blr_start_master_in_main,
router,
connect_retry);
MXS_FREE(name);
MXS_ERROR("%s: failed to connect to master server '%s', "
"retrying in %d seconds",
router->service->name,
router->service->dbref->server->name,
connect_retry);
}
MXS_ERROR("%s: failed to connect to master server '%s', "
"retrying in %d seconds",
router->service->name,
router->service->dbref->server->name,
connect_retry);
}
else
{
@ -489,29 +465,6 @@ blr_master_close(ROUTER_INSTANCE *router)
router->stored_event = NULL;
}
/**
* Mark this master connection for a delayed reconnect, used during
* error recovery to cause a reconnect after router->retry_interval seconds.
*
* @param router The router instance
*/
void
blr_master_delayed_connect(ROUTER_INSTANCE *router)
{
static const char master[] = "Master Recovery";
char *name = (char *)MXS_MALLOC(strlen(router->service->name) + sizeof(master));
if (name)
{
sprintf(name, "%s %s", router->service->name, master);
hktask_add(name,
blr_start_master_in_main,
router,
router->retry_interval);
MXS_FREE(name);
}
}
/**
* Binlog router master side state machine event handler.
*
@ -1084,7 +1037,7 @@ blr_handle_binlog_record(ROUTER_INSTANCE *router, GWBUF *pkt)
{
MXS_FREE(msg);
blr_master_close(router);
blr_master_delayed_connect(router);
blr_start_master_in_main(router);
return;
}