centos6.8 x86_64,

svoj, marko, Elkin, thoughts?

This is wide, can be locking, can be innodb, can even be related to semi-sync somehow, I guess.

FWICS clients connect from different host. According to the "stack" most threads stuck in poll/select. Can this be due to network problem?

Who could say how handler::ha_rnd_next() ends up in calling select ():

#0  0x00007f6dd44a6623 in select () from /lib64/libc.so.6

#1  0x00000000009b9f0f in ?? ()

#2  0x0000000000a83404 in ?? ()

#3  0x0000000000a749cd in ?? ()

#4  0x0000000000a8670c in ?? ()

#5  0x0000000000a87163 in ?? ()

#6  0x0000000000a73903 in ?? ()

#7  0x0000000000a616e5 in ?? ()

#8  0x0000000000a07c21 in ?? ()

#9  0x0000000000a0a8e0 in ?? ()

#10 0x000000000096d4cb in ?? ()

#11 0x00000000006ea044 in handler::ha_rnd_next(unsigned char*) ()

#12 0x00000000007d7f3c in rr_sequential(READ_RECORD*) ()

#13 0x00000000007f4914 in mysql_delete(THD*, TABLE_LIST*, Item*, SQL_I_List<st_order>*, unsigned long long, unsigned long long, select_result*) ()

#14 0x0000000000598eb2 in mysql_execute_command(THD*) ()

#15 0x000000000059c921 in mysql_parse(THD*, char*, unsigned int, Parser_state*) ()

As to the semisync involvement, messages like

180427 12:48:21 [Warning] Timeout waiting for reply of binlog (file: SGRDB-mysql01-bin.010453, pos: 520), semi-sync up to

appear in the log sporadically all the time, but there was always progress. I would downvote it as the main suspect.

First i guess select() in handler::ha_rnd_next is the behavior of thread-pool, but check the configure file, not enabled thread pool. Maybe Async IO?

louis.hust,

Are you still experiencing the problem?
It appears that you're building the binaries from source. Can you resolve the stack trace?

hi, @elenst

Not experiencing the problem.

The stack file is in attachments. How to resolve it?

You would need the same exact binary, but if you're not having the problem any longer, you've probably upgraded since then.
Here is another question – your error log shows that you have CONNECT engine installed, could this stack trace actually belong to it and not to plain InnoDB?

The same exact binary means what?
Never used connect engine.

The same exact mysqld binary which you used at the time when the stack trace was produced. Are you still running it? If you are, have you done something to make the problem disappear? Or was it a one-time occurrence?