xtrabackup SST : I/O error reading the header from the binary log

1,390 views
Skip to first unread message

Florent B

unread,
Mar 17, 2015, 12:47:09 PM3/17/15
to codersh...@googlegroups.com
Hi,

I use Mariadb-galera-server 10 packaged for Debian.

I want to setup a new cluster, all works fine with rsync SST method.

But when I switch to xtrabackup or xtrabackup-v2 (and sst_auth), new
nodes joining the cluster fail to start with:

mysqld: 150317 17:05:10 [ERROR] Can't init tc log
...
mysqld: 150317 17:05:10 [ERROR] I/O error reading the header from the
binary log, errno=175, io cache code=0
...
mysqld: 150317 17:05:10 [ERROR] I/O error reading the header from the
binary log
...
mysqld: 150317 17:05:10 [ERROR] Aborting
...

I think SST is successful because I have this in log:
mysqld: 150317 17:04:56 [Note] WSREP: 1.0 (staging1): State transfer to
0.0 (staging3) complete.
...
mysqld: 150317 17:05:08 [Note] WSREP: SST complete, seqno: 1


/var/log/mysql/mariadb-bin.state exists but is empty (last file readed
before error in strace)
/var/log/mysql/mariadb-bin.index contains "/var/log/mysql/mariadb-bin.state"
(no other file in /var/log/mysql on joining node)

This is a freshly created cluster, without any client using it.

What could cause that ?

I tried multiple times, re-starting from OS install, and always the same
error...

Let me known if you need some other information.

Thank you.

Gavin Alves

unread,
Mar 29, 2015, 11:02:40 AM3/29/15
to codersh...@googlegroups.com
I have the exact same problem.

Joffrey MICHAÏE

unread,
Mar 29, 2015, 1:19:14 PM3/29/15
to Gavin Alves, codersh...@googlegroups.com
Can you try disabling the binary log on the joiner (for the first sst) ? There may be a problem on that side.

Joffrey


--
You received this message because you are subscribed to the Google Groups "codership" group.
To unsubscribe from this group and stop receiving emails from it, send an email to codership-tea...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

esp3482

unread,
May 5, 2015, 8:55:35 AM5/5/15
to codersh...@googlegroups.com, ad...@mediafire.co.uk
Hi,

I also had the same problem (Debian 7, mariadb-galera-server 10.0.17, percona-xtrabackup 2.2.10)

Disabling the binary log worked. After the node successfully joined, binary logging can be re-enabled.

Couple of weeks ago I build another test cluster with mariadb-galera-server version 10.0.15 without this issue, is this possible?

Daniel

Gavin Alves

unread,
May 5, 2015, 8:07:29 PM5/5/15
to codersh...@googlegroups.com, ad...@mediafire.co.uk
I believe this is caused by a bug when using a non-default log_bin location on the joiner.

Set:

log_bin=binlog

and retry.

esp3482

unread,
May 13, 2015, 5:32:16 AM5/13/15
to codersh...@googlegroups.com
Hi Gavin,

I'm facing another issue with joining a new node right now

https://groups.google.com/forum/?hl=de#!topic/codership-team/PHwTY27oqxU

But I will test it, if the error appears again and come back to you..

Thanks
Daniel
Message has been deleted

esp3482

unread,
May 13, 2015, 9:35:16 AM5/13/15
to codersh...@googlegroups.com, ad...@mediafire.co.uk
Hi,

did you take a look in your my.cnf? In my debian/mariadb system the default-value is:

log_bin = /var/log/mysql/mariadb-bin


Am Mittwoch, 13. Mai 2015 14:49:32 UTC+2 schrieb Florent B:
Hi,

I do not redefine "log_bin" option in my config, so it is set at its default value I think...


Message has been deleted

esp3482

unread,
May 14, 2015, 7:54:46 AM5/14/15
to codersh...@googlegroups.com, ad...@mediafire.co.uk
Hi,

like I said, I didn't had this issue with galera 10.0.15, maybe it's new in 10.0.17?

Did you try to set it explicitly to "log_bin=binlog" like Gavin said? I still cant't test it myself becaue of another error...



Am Donnerstag, 14. Mai 2015 13:27:18 UTC+2 schrieb Florent B:
Hi,

Yes I have the same default value of course.
Why MariaDB is setting this as default value if it does not work with xtrabackup SST ?


esp3482

unread,
May 18, 2015, 9:34:46 AM5/18/15
to codersh...@googlegroups.com
Hi Gavin,

i installed another test cluster and used for the joiner "log_bin=binlog"

#log_bin = /var/log/mysql/mariadb-bin
#log_bin_index = /var/log/mysql/mariadb-bin.index
log_bin = binlog

Joining does work now... if "binlog" is the internal default value it's no wonder that "deactivating" binary log by using

#log_bin = /var/log/mysql/mariadb-bin
#log_bin_index = /var/log/mysql/mariadb-bin.index

had worked to in my previous test...

Is there a way to fix this issue? When the node has finally joined, you can active binary logging to the non-default-values and start the daemon again and everything works fine...

Thanks
Daniel
Reply all
Reply to author
Forward
0 new messages