[prev in list] [next in list] [prev in thread] [next in thread] 

List:       slony1-general
Subject:    Re: [Slony1-general] Slony 2.2.3 extended lag recovery
From:       Glyn Astill <glynastill () yahoo ! co ! uk>
Date:       2015-02-02 17:59:54
Message-ID: 725041738.1147357.1422899994574.JavaMail.yahoo () mail ! yahoo ! com
[Download RAW message or body]

----- Original Message -----

> From: Steve Singer <ssinger@ca.afilias.info>
> To: slony1-general@lists.slony.info
> Cc: 
> Sent: Monday, 2 February 2015, 14:41
> Subject: Re: [Slony1-general] Slony 2.2.3 extended lag recovery
> 
> On 01/29/2015 12:01 PM, Glyn Astill wrote:
> 
> 
> Why are you regularly running vacuum full (versus normal vaccum?)

> 

We're not regularly running vacuum full, I didn't state that.

> Vacuum full will take an exclusive lock on the table for its duration. 
> Also the cluster command tends to be faster than vacuum full but with 
> similar results (re-writing your tables)
> 
> Are you running vacuum full on specific tables or on all tables in your 
> cluster?

> 

Understood.

Just the one, just the once, I only mentioned it because it was the thing that caused \
the lag build up - and the lag was expected, no ignorance there.


However we've had similar long running backlogs since moving to 2.2 and we're \
definatey not ever running vacuum full at any other time.

> If you do a vacuum full on a table on your replica slony won't be able 
> to insert any data into that table until the vacuum transaction is 
> complete and the exclusive lock is released.
> 
> 


Understood.  My point is that we seem to be seeing lag recovery taking longer than \
expected since slony switched to using copy vs regular insert/update/delete.  


From what I'm seeing, there's some condition where the copy process stalls or starts \
doing something strange.  In this case it was an exclusive lock that lasted 2 \
minutes, causing a backlog that took over 20 minutes to apply, where I'm prety \
confident at our transaction rate that if I'd stoped the slons and restarted them 2 \
minutes later we'd have caught up in less than 2 minutes. \
_______________________________________________ Slony1-general mailing list
Slony1-general@lists.slony.info
http://lists.slony.info/mailman/listinfo/slony1-general


[prev in list] [next in list] [prev in thread] [next in thread] 

Configure | About | News | Add a list | Sponsored by KoreLogic