garbd crashes on joining

[Vadim Tkachenko]

garbd from galera/2.x crashes

./garbd --address=gcomm://10.11.12.206 --group=big_monkey
 INFO: Read config:
        daemon: 0
        address: gcomm://10.11.12.206
        group: big_monkey
        sst: none
        donor:
        options: gcs.fc_limit=9999999; gcs.fc_factor=1.0; gcs.fc_master_slave=yes
        cfg:
        log:

 INFO: protonet asio version 0
 INFO: backend: asio
 INFO: GMCast version 0
 INFO: (f99c61b4-1001-11e1-0800-4cc9258aa7f5, 'tcp://0.0.0.0:4567') listening at tcp://0.0.0.0:4567
 INFO: (f99c61b4-1001-11e1-0800-4cc9258aa7f5, 'tcp://0.0.0.0:4567') multicast: , ttl: 1
 INFO: EVS version 0
 INFO: PC version 0
 INFO: gcomm: connecting to group 'big_monkey', peer '10.11.12.206:'
 INFO: GMCast::handle_stable_view: view(view_id(PRIM,838e1b60-1003-11e1-0800-31e59e3e0a8f,2) memb {
        838e1b60-1003-11e1-0800-31e59e3e0a8f,
        f99c61b4-1001-11e1-0800-4cc9258aa7f5,
} joined {
} left {
} partitioned {
})
 INFO: declaring 838e1b60-1003-11e1-0800-31e59e3e0a8f stable
 INFO: gcomm: connected
 INFO: Changing maximum message size 0 -> 64500, result: 32636
 INFO: Shifting CLOSED -> OPEN (TO: 0)
 INFO: Opened channel 'big_monkey'
 INFO: New COMPONENT: primary = yes, my_idx = 1, memb_num = 2
 INFO: STATE EXCHANGE: Waiting for state UUID.
 INFO: STATE EXCHANGE: sent state msg: a2545b94-1003-11e1-0800-5788984b3507
 INFO: STATE EXCHANGE: got state msg: a2545b94-1003-11e1-0800-5788984b3507 from 0 (hppro1.office.percona.com)
 INFO: STATE EXCHANGE: got state msg: a2545b94-1003-11e1-0800-5788984b3507 from 1 (garb)
 INFO: Quorum results:
        version = 1,
        component = PRIMARY,
        conf_id = 1,
        members = 1/2 (joined/total),
        act_id = 606442,
        last_appl. = -1,
        protocols = 0/1/1 (gcs/repl/appl),
        group UUID = e847fdec-0fbb-11e1-0800-86297d0ef738
 INFO: Flow-control interval: [9999999, 9999999]
 INFO: Shifting OPEN -> PRIMARY (TO: 606442)
 INFO: Sending state transfer request: 'none', size: 4
Segmentation fault (core dumped)

[Vadim Tkachenko (vadim-tk)]

Backtrace is
(gdb) bt
#0 0x000000373312686f in __strlen_sse42 () from /lib64/libc.so.6
#1 0x0000000000422311 in gcs_group_handle_state_request (group=0x74ed60, act=0x7ffff747ed30) at gcs/src/gcs_group.c:907
#2 0x000000000041bea2 in core_handle_act_msg (conn=0x74ecd0, recv_act=0x7ffff747ed30, timeout=9223372035999999999)
    at gcs/src/gcs_core.c:547
#3 gcs_core_recv (conn=0x74ecd0, recv_act=0x7ffff747ed30, timeout=9223372035999999999) at gcs/src/gcs_core.c:997
#4 0x0000000000418ce0 in gcs_recv_thread (arg=0x74e890) at gcs/src/gcs.c:1108
#5 0x00000037338077e1 in start_thread () from /lib64/libpthread.so.0
#6 0x00000037330e68ed in clone () from /lib64/libc.so.6

Because of this crash I can't test 2 node + garbd ( from 2.x) configutation

[Alex Yurchenko (ayurchen)]

fix committed in r99