On 07/29/2015 09:34 AM, David Fetter wrote:
> On Tue, Jul 28, 2015 at 06:51:44PM -0400, Jan Wieck wrote:
>> On 07/25/2015 12:31 AM, David Fetter wrote:
>> >Folks,
>> >
>> >While in the best of all possible worlds, we'd have planned out a
>> >replication strategy before we get tables whose initial sync via
>> >"SUBSCRIBE SET" will never finish, we aren't always given that much
>> >ability to plan that soon.
>>
>> One thing I forgot to ask initially. What do you mean by "will never
>> finish"? Is that just being facetious about "it will take a long
>> time", is it that you don't have the disk storage to swallow the
>> back log that will accumulate or is it "outdated knowledge" from the
>> times when Slony did actually slow down with accumulation of
>> backlog?
>
> As far as I could tell, backlog was accumulating faster than the
> initial sync was happening.

If that is the case with 2.2, then there is a good chance that the slave 
won't be able to keep up anyway.

Note that the log switching on the master is severely hindered until the 
slave is actually caught up. So you will see a sawtooth pattern in the 
total amount of sl_log. I can explain that in more detail if needed.

Assuming a more or less steady rate of update activity on the master, 
the rate at which SYNC events get replicated is a better indicator. If 
master-SYNC event appear on the slave faster than they are generated on 
the master, it is catching up.


Regards, Jan

-- 
Jan Wieck
Senior Software Engineer
http://slony.info