I meet the core ,use librdkafka-0.8.6.zip and Linux version 2.6.32, gcc version 4.8.2

[yangkai04]

Program terminated with signal 11, Segmentation fault.
#0 0x00000000004c61d9 in RdKafka::log_cb_trampoline (rk=, level=3, fac=,

buf=0x7ff58f39a620 "nmg01-crm-hina59.nmg01:8092/59: Receive failed: Receive error: Connection timed out") at HandleImpl.cpp:55

55 handle->event_cb_->event_cb(event);
(gdb) bt
#0 0x00000000004c61d9 in RdKafka::log_cb_trampoline (rk=, level=3, fac=,

buf=0x7ff58f39a620 "nmg01-crm-hina59.nmg01.baidu.com:8092/59: Receive failed: Receive error: Connection timed out") at HandleImpl.cpp:55

#1 0x00000000004c7f69 in rd_kafka_log0 (rk=0x22be000, extra=extra@entry=0x0, level=level@entry=3, fac=fac@entry=0x6950ac "FAIL", fmt=fmt@entry=0x6951d5 "%s")

at rdkafka.c:141

#2 0x00000000004d58ce in rd_kafka_broker_fail (rkb=rkb@entry=0x2391800, err=RD_KAFKA_RESP_ERR__TRANSPORT, fmt=fmt@entry=0x69511d "Receive failed: %s")

at rdkafka_broker.c:398

#3 0x00000000004d5fe7 in rd_kafka_recv (rkb=rkb@entry=0x2391800) at rdkafka_broker.c:1519
#4 0x00000000004d6888 in rd_kafka_broker_io_serve (rkb=rkb@entry=0x2391800) at rdkafka_broker.c:2452
#5 0x00000000004d7103 in rd_kafka_broker_consumer_serve (rkb=0x2391800) at rdkafka_broker.c:4105
#6 rd_kafka_broker_thread_main (arg=) at rdkafka_broker.c:4154
#7 0x00007ff5a07b71c3 in ?? ()
#8 0x0000000000000000 in ?? ()

[edenhill]

Are you getting this after destroying a Kafka handle?

[edenhill]

Can do this in gdb?

 frame 0
 p *handle
 p *handle->event_cb_
 p *event

[yangkai04]

not destroy , when i get the core , it is running, ok, wait for a moment

[yangkai04]

Program terminated with signal 11, Segmentation fault.
#0 0x00000000004c61d9 in RdKafka::log_cb_trampoline (rk=, level=3, fac=,
buf=0x7ff58f39a620 "nmg01-crm-hina59.nmg01:8092/59: Receive failed: Receive error: Connection timed out") at HandleImpl.cpp:55
55 handle->event_cb_->event_cb(event);
(gdb) p handle
$1 = {RdKafka::Handle = {vptr.Handle = 0x98d6e0 <vtable for RdKafka::ConsumerImpl+96>}, rk = 0x22be000, event_cb = 0x7fff3bd57db0, socket_cb_ = 0x0,
open_cb_ = 0x0, dr_cb_ = 0x0, partitioner_cb_ = 0x0}
(gdb) p *handle->event_cb_
$2 = {vptr.EventCb = 0x9}
(gdb) p *event
No symbol "operator" in current context.

[yangkai04]

when init, I did follows:
_consumer = RdKafka::Consumer::create(_conf, errstr);
_tconf = RdKafka::Conf::create(RdKafka::Conf::CONF_TOPIC);
_topic = RdKafka::Topic::create(_consumer, _topicname, _tconf, errstr);

// because there is 12 partition, so I start 12 partid
for (int partid = 0; partid < _partition; ++partid) {
RdKafka::ErrorCode resp = _consumer->start(_topic, partid, start_offset);
EXPECT_EQ_OR_RETURN_FAIL_WITH_MSG(resp, RdKafka::ERR_NO_ERROR, RdKafka::err2str(resp));
}

when run, i did:
for (int partid = 0; partid < _partition; ++partid) {
RdKafka::Message *msg = _consumer->consume(_topic, partid, 1000);
...
}

if I use with some wrong?

[yangkai04]

I can recieve msg, but when run for a while, I meet the problem.

it seems a kafka broker down or some other problems caused.

[edenhill]

How did you set up _conf?

[yangkai04]

set up _conf as follows:
_conf = RdKafka::Conf::create(RdKafka::Conf::CONF_GLOBAL);
res = _conf->set("metadata.broker.list", _brokers, errstr);
ExampleEventCb ex_event_cb;
res = _conf->set("event_cb", &ex_event_cb, errstr);

[yangkai04]

class ExampleEventCb : public RdKafka::EventCb {
public:
void event_cb (RdKafka::Event &event) {
switch (event.type()) {
case RdKafka::Event::EVENT_ERROR:
FATAL_LOG("ERROR ( %s ): %s",
RdKafka::err2str(event.err()).c_str(), event.str().c_str());
if (event.err() == RdKafka::ERR__ALL_BROKERS_DOWN) {
FATAL_LOG("ERROR ( %s ): %s",
RdKafka::err2str(event.err()).c_str(), event.str().c_str());
}
break;
case RdKafka::Event::EVENT_STATS:
NOTICE_LOG("starts %s", event.str().c_str());
break;
case RdKafka::Event::EVENT_LOG:
FATAL_LOG("LOG-%i-%s: %s", event.severity(), event.fac().c_str(), event.str().c_str());
break;
default:
FATAL_LOG("EVENT type(%s) ( %s ): %s",
event.type(), RdKafka::err2str(event.err()).c_str(), event.str().c_str());
break;
}
}
};

[yangkai04]

by the way, when meet net error or other error, need i restart the consumer? what error need to restart and how?

thank you~!

[edenhill]

I'm guessing your EventCb is stack allocated and goes out of scope.

[edenhill]

You do not need to take any action when broker connection goes down, librdkafka will reconnect automatically.