Percona XtraDB Cluster moved to https://jira.percona.com/projects/PXC

  1. Bug #1197771
  2. Comment #29

Comment 29 for bug 1197771

Revision history for this message
Dmitry Yu Okunev (dyokunev) wrote :

I have a similar issue with different versions of galera/mysql and OS:
The working node have a lot of "wsrep in pre-commit stage" processes (without any CPU/HDD/network load), but the receiving node have a process in stage of "System lock" with 100% CPU load.

I've tried to attach with gdb on the moment of problem and got this backtrace:

#0 0x00007fa4c1bf34a3 in poll () from /lib/x86_64-linux-gnu/libc.so.6
#1 0x00000000005f92b2 in handle_connections_sockets () at /mnt/workspace/percona-xtradb-cluster-5.6-debs/label_exp/debian-wheezy-x64/target/Percona-XtraDB-Cluster-5.6.15/sql/mysqld.cc:7099
#2 0x00000000006001f6 in mysqld_main (argc=68, argv=0x137f028) at /mnt/workspace/percona-xtradb-cluster-5.6-debs/label_exp/debian-wheezy-x64/target/Percona-XtraDB-Cluster-5.6.15/sql/mysqld.cc:6494
#3 0x00007fa4c1b40ead in __libc_start_main () from /lib/x86_64-linux-gnu/libc.so.6
#4 0x00000000005f30fd in _start ()

I'll try to recompile with without "HAVE_POLL", now.