[Bucardo-general] Recover bucardo "hole" in sync

[Laurent GARCES]

Hi,

Thanks for the answer. It's not exactly what we are looking for.

Our goal is to "double check" that bucardo is doing is job right. We use 
bucardo for some years now and never really had trouble, but things can 
always goes wrong...
We already monitor every few minutes that the result of "bucardo status" 
is correct. But we would like to directly check that the data are in 
sync without using any bucardo functionality/log, as the goal is to 
verify that there is no (silent) problem in bucardo. So using only 
SQL/Postgres tools. As there is a lot of activity in our new project 
(thousands "write" requests by hour) we cannot check the whole data 
everytime. The idea would be to verify, each day for example, that the 
data content that have been modified the day before are in sync. We 
don't know if it's possible without using any "timestamp information" in 
the data itself but may there are some "tricks" we are not aware that 
could do the job.
I hope I'm clear enough.

Thanks for your help,

Regards,

-- 
Laurent GARCES
LRD www.l-rd.fr
+33 (0)4 67 92 52 56

Le 19/01/2017 à 16:57, Greg Sabino Mullane a écrit :
> On Mon, Jan 16, 2017 at 12:34:37PM +0100, Laurent GARCES wrote:
> ...
>> As a side question, could you advice a way to monitor the bucardo job on
>> postgres?
>>
>> During the tests phase we did some things like, on each synced database:
>>
>> pg_dump <database> --exclude-schema=bucardo -t <syncedTable1> -t
>> <syncTable2> | sort | md5sum
> ...
>
> So you are looking for a list of tables that have changed (via Bucardo)
> during a certain period of time? Parsing of the log.bucardo file may
> be an option. Make sure log_level = verbose (or debug if you are brave
> and have disk space to burn :) and you can grep for entries like this:
>
> (13639) [Fri Jan 13 17:38:10.588 2017] #10014 KID (mongo) Rows copied to (postgres) C.public.bucardo_test9: 1
>
> If you can describe what you are looking for in more detail, there may be
> some different options, but that's the first thing that springs to mind.
>