[Bucardo-general] Latency on delete

Alex Balashov abalashov at evaristesys.com
Fri Jan 17 16:27:39 UTC 2014


On 01/17/2014 11:09 AM, Greg Sabino Mullane wrote:

> On Fri, Jan 17, 2014 at 10:15:20AM -0500, Alex Balashov wrote:
>> How long is a DELETE of ~4.5 million rows (in a relatively narrow
>> table) supposed to take in a scenario with a simple pushdelta
>> replication setup?
>
> That's a lot of rows. Your best bet is to make sure you are logging
> long queries at the Postgres level (e.g. setting log_min_duration_statment = 500),
> and then see the exact queries that are taking up most of the time.

Yeah, I hear you.  In fact, I'm not even sure this is a Bucardo issue 
per se, and I would expect any deletion of so many rows to churn for a 
while.

But still, am I right to assume that the deletion of any given row won't 
return, as a statement, until the trigger that pushes that delta into 
the Bucardo tables returns?  It's a fairly "synchronous" process, right?

I haven't tried that deletion with the Bucardo delta triggers disabled, 
and given that this is a production system, it'd be kind of challenging 
to do that.  But I agree with you that slow statement logs would 
probably illuminate the problem better, and will go down that avenue.

> If you are using the latest Bucardo and going to multiple targets,
> I just pushed an optimization that should help that case, so git pull
> and try again :)

I'm not, I'm using the 4.5.0 tarball from the web site.  What's the 
optimisation?

-- Alex

-- 
Alex Balashov - Principal
Evariste Systems LLC
235 E Ponce de Leon Ave
Suite 106
Decatur, GA 30030
United States
Tel: +1-678-954-0670
Web: http://www.evaristesys.com/, http://www.alexbalashov.com/


More information about the Bucardo-general mailing list