Web lists-archives.com

Re: io thread very slow copying binlogs




I should add that I've turned the SQL thread off, it makes no difference
from what I can see...

On Fri, 20 Jan 2017 10:31:38 -0500
Brad Barnett <bahb@xxxxxxx> wrote:

> 
> 
> 
> On Fri, 20 Jan 2017 10:30:09 -0500
> Brad Barnett <mysql-general-list@xxxxxxx> wrote:
> 
> > 
> > 
> > Hey Morgan,
> > 
> > Thanks for the tip.  Might come in handy.
> > 
> > But, I'm positive it's not a disconnect / reconnect thing.  Or, at
> > least not one affected by that timeout.
> > 
> > I can do a watch ls -lh in the binlog dir, and see the relay log
> > increasing in size by a M every 4 or 5 seconds or so.  About
> > 200kbyte/sec / 1.6Mbit/sec right now.
> > 
> > It seems very steady too.  As in, if I look at bytes, they're
> > constantly increasing.. just, slow..
> > 
> > On Fri, 20 Jan 2017 10:19:57 -0500
> > "Morgan Tocker" <morgan.tocker@xxxxxxxxxx> wrote:
> > 
> > > Hi Brad,
> > > 
> > > > MySQL community edition 5.6.29, running Linux.
> > > > 
> > > > Binlogs never seem to get caught up on slaves.
> > > > 
> > > > I've done all I can, to validate that this isn't network or disk
> > > > related.
> > > > 
> > > > Disk tests (using iostat and other methods) show lots of bandwidth
> > > > left on the slave and master.
> > > > 
> > > > Network tests, such as:
> > > > 
> > > > - using scp to copy binlogs directly
> > > > - using different NICs to copy binlogs
> > > > - using mysqlbinlog to snag logs (the most 'real' way I can think
> > > > to simulate the replication thread copying binlogs from the
> > > > master)
> > > > 
> > > > All seem to show that network speed is blazingly fast.
> > > > 
> > > > Yet, MySQL is barely getting 4mbit/sec across the network, and
> > > > onto the
> > > disk.
> > > > And that's on a good day.
> > > > 
> > > > Any immediate suggestions here?  This seems very weird, and SQL
> > > > thread is constantly running out of stuff to process.
> > > 
> > > Networking is not my strong-suit, but I have a suggestion:
> > > 
> > > Try lowering slave-net-timeout
> > > http://dev.mysql.com/doc/refman/5.6/en/replication-options-slave.html#option
> > > _mysqld_slave-net-timeout
> > > 
> > > I remember that we lowered the default in MySQL 5.7 (from 1hr to 60
> > > seconds) so that the connection between master/slave would be
> > > considered broken faster.  If you have the throughput on a graph it
> > > might better explain if it is a constant 4mbit/sec or more broken.
> > > 
> > > 
> > > - Morgan

-- 
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/mysql