sin/samba-mirror
1
0
Fork 0
mirror of https://github.com/samba-team/samba.git synced 2025-01-10 01:18:15 +03:00
Code

if the event scripts hangs EventScriptsBanCount consecutive times in a row

Browse Source 
the node will ban itself for the default recovery ban period

(This used to be ctdb commit 7239d7ecd54037b11eddf47328a3129d281e7d4a)
This commit is contained in:
Ronnie Sahlberg 2008-06-13 13:18:06 +10:00
parent 30535c815d
commit 779468ab3f
3 changed files with 34 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ctdb/include/ctdb_private.h
View File

@ -102,6 +102,7 @@ struct ctdb_tunable {
uint32_t monitor_interval; uint32_t monitor_interval;
uint32_t tickle_update_interval; uint32_t tickle_update_interval;
uint32_t script_timeout; uint32_t script_timeout;
uint32_t script_ban_count; /* ban after this many consec timeouts*/
uint32_t recovery_grace_period; uint32_t recovery_grace_period;
uint32_t recovery_ban_period; uint32_t recovery_ban_period;
uint32_t database_hash_size; uint32_t database_hash_size;
@ -410,6 +411,7 @@ struct ctdb_context {
struct ctdb_monitor_state *monitor; struct ctdb_monitor_state *monitor;
struct ctdb_log_state *log; struct ctdb_log_state *log;
int start_as_disabled; int start_as_disabled;
uint32_t event_script_timeouts; /* counting how many consecutive times an eventscript has timedout */
TALLOC_CTX *eventscripts_ctx; /* a context to hold data for the RUN_EVENTSCRIPTS control */ TALLOC_CTX *eventscripts_ctx; /* a context to hold data for the RUN_EVENTSCRIPTS control */
}; };

1
ctdb/server/ctdb_tunables.c
View File

@ -38,6 +38,7 @@ static const struct {
{ "MonitorInterval", 15, offsetof(struct ctdb_tunable, monitor_interval) }, { "MonitorInterval", 15, offsetof(struct ctdb_tunable, monitor_interval) },
{ "TickleUpdateInterval",20, offsetof(struct ctdb_tunable, tickle_update_interval) }, { "TickleUpdateInterval",20, offsetof(struct ctdb_tunable, tickle_update_interval) },
{ "EventScriptTimeout", 20, offsetof(struct ctdb_tunable, script_timeout) }, { "EventScriptTimeout", 20, offsetof(struct ctdb_tunable, script_timeout) },
{ "EventScriptBanCount", 5, offsetof(struct ctdb_tunable, script_ban_count) },
{ "RecoveryGracePeriod", 60, offsetof(struct ctdb_tunable, recovery_grace_period) }, { "RecoveryGracePeriod", 60, offsetof(struct ctdb_tunable, recovery_grace_period) },
{ "RecoveryBanPeriod", 300, offsetof(struct ctdb_tunable, recovery_ban_period) }, { "RecoveryBanPeriod", 300, offsetof(struct ctdb_tunable, recovery_ban_period) },
{ "DatabaseHashSize", 10000, offsetof(struct ctdb_tunable, database_hash_size) }, { "DatabaseHashSize", 10000, offsetof(struct ctdb_tunable, database_hash_size) },

33
ctdb/server/eventscript.c
View File

@ -222,6 +222,27 @@ static void ctdb_event_script_handler(struct event_context *ev, struct fd_event
talloc_set_destructor(state, NULL); talloc_set_destructor(state, NULL);
talloc_free(state); talloc_free(state);
callback(ctdb, status, private_data); callback(ctdb, status, private_data);
ctdb->event_script_timeouts = 0;
}
static void ctdb_ban_self(struct ctdb_context *ctdb, uint32_t ban_period)
{
int ret;
struct ctdb_ban_info b;
TDB_DATA data;
b.pnn = ctdb->pnn;
b.ban_time = ban_period;
data.dptr = (uint8_t *)&b;
data.dsize = sizeof(b);
ret = ctdb_daemon_send_message(ctdb, CTDB_BROADCAST_CONNECTED,
CTDB_SRVID_BAN_NODE, data);
if (ret != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to send ban message\n"));
}
} }
@ -234,9 +255,17 @@ static void ctdb_event_script_timeout(struct event_context *ev, struct timed_eve
void *private_data = state->private_data; void *private_data = state->private_data;
struct ctdb_context *ctdb = state->ctdb; struct ctdb_context *ctdb = state->ctdb;
DEBUG(DEBUG_ERR,("event script timed out : %s\n", state->options)); DEBUG(DEBUG_ERR,("Event script timed out : %s count : %u\n", state->options, ctdb->event_script_timeouts));
talloc_free(state); talloc_free(state);
callback(ctdb, -1, private_data); callback(ctdb, -1, private_data);
ctdb->event_script_timeouts++;
if (ctdb->event_script_timeouts > ctdb->tunable.script_ban_count) {
ctdb->event_script_timeouts = 0;
DEBUG(DEBUG_ERR, ("Maximum timeout count reached for eventscript. Banning self for %d seconds\n", ctdb->tunable.recovery_ban_period));
ctdb_ban_self(ctdb, ctdb->tunable.recovery_ban_period);
}
} }
/* /*
@ -308,7 +337,7 @@ static int ctdb_event_script_callback_v(struct ctdb_context *ctdb,
if (!timeval_is_zero(&timeout)) { if (!timeval_is_zero(&timeout)) {
event_add_timed(ctdb->ev, state, timeout, ctdb_event_script_timeout, state); event_add_timed(ctdb->ev, state, timeout, ctdb_event_script_timeout, state);
} else { } else {
DEBUG(DEBUG_ERR, (__location__ " eventscript %s called with no timeout\n", fmt)); DEBUG(DEBUG_ERR, (__location__ " eventscript %s called with no timeout\n", state->options));
} }
return 0; return 0;