GTM standby crashes in the normal operation. Cluster consists of two gtm proxies, two datanodes and two coordinators. The operation is as follows:
i) Initialize each coordinator/datanode with initidb,
ii) Edit configuration files (gtm, gtm_standby, gtm_proxy, coordinators, datanodes) manually (in fact, using dedicated script)
iii) Start all the gtms
iv) Create NODE, create database
v) sql -c 'select 1'
The backtrace looks like:
--------------
(gdb) bt
#0 0x0000000000409645 in GTM_BkupBeginTransactionGetGXIDMulti (
coord_name=0x434f41 "", txn=0x7fcd4e914bc0, gxid=0x7fcd4e910bc0,
isolevel=0x7fcd4e90cbc0, readonly=0x7fcd4e91abd0 "",
connid=0x7fcd4e918bc0, txn_count=2) at gtm_txn.c:1230
#1 0x000000000040a3f3 in ProcessBkupBeginTransactionGetGXIDCommandMulti (
myport=0x86ce70, message=0x7fcd4e91bd70) at gtm_txn.c:1539
#2 0x0000000000405ee0 in ProcessTransactionCommand (myport=0x86ce70,
mtype=MSG_BKUP_TXN_BEGIN_GETGXID_MULTI, message=0x7fcd4e91bd70)
at main.c:1487
#3 0x0000000000405966 in ProcessCommand (myport=0x86ce70,
input_message=0x7fcd4e91bd70) at main.c:1238
#4 0x0000000000405748 in GTM_ThreadMain (argp=0x871020) at main.c:1112
#5 0x00000000004074f2 in GTM_ThreadMainWrapper (argp=0x871020)
at gtm_thread.c:343
#6 0x00007fcd5016c9ca in start_thread (arg=<value optimized out>)
at pthread_create.c:300
#7 0x00007fcd4fec9cdd in clone ()
at ../sysdeps/unix/sysv/linux/x86_64/clone.S:112
#8 0x0000000000000000 in ?? ()
(gdb)
--------------
This might have been fixed by the series of patch by Andrei. Thanks Andrei!! We should continue to see if the similar problem occurs in the future.