Re: [BUGS] BUG #14530: Logical Decoding Slowness

On 2017-02-09 06:55:00 +0000, Huan Ruan wrote:
> Hi Andres
> 
> 
> > Unfortunately perf files can't be interpreted on other systems unless
> > you added all the referenced object files to it using perf archive.
> >
> 
> Sorry this was my first time using perf for profiling so didn't know that.
> Also, the profile was collected without Postgres debuginfo installed. If
> you need that or something else please let me know.
> 
> 
> > Could you perhaps just send the first few lines of the report?
> >
> > The first screen of 'perf report -n' for subtransaction case is,
> 
> Samples: 247K of event 'cycles', Event count (approx.): 155149593065
>   Children      Self       Samples  Command          Shared Object
>       Symbol
> +   51.26%     0.00%             0  postgres         [unknown]
>       [.] 0x0000000003236e68
> +   48.66%    48.54%        117427  postgres         postgres
>      [.] CatalogCacheIdInvalidate
> +   38.44%    38.35%         92748  postgres         postgres
>      [.] hash_seq_search
> +    5.67%     5.66%         13683  postgres         postgres
>      [.] CallSyscacheCallbacks
> +    3.00%     2.99%          7234  postgres         postgres
>      [.] LocalExecuteInvalidationMessage
> +    1.00%     0.99%          2397  postgres         postgres

Yea, that was what I was thinking - we're replaying more invalidations
than we'd need to, and we do so very frequently.  That's quite
expensive.  Hmmm.

> for create tables is,
> 
> Samples: 342K of event 'cycles', Event count (approx.): 216348097599
>   Children      Self       Samples  Command          Shared Object
>       Symbol
> +   52.09%     0.00%             0  postgres         [unknown]
>       [.] 0x000000009754f3a8
> +   49.09%    48.98%        164831  postgres         postgres
>      [.] CatalogCacheIdInvalidate
> +   38.75%    38.67%        130091  postgres         postgres
>      [.] hash_seq_search
> +    4.92%     4.90%         16494  postgres         postgres
>      [.] CallSyscacheCallbacks
> +    3.20%     3.19%         10736  postgres         postgres
>      [.] LocalExecuteInvalidationMessage

Same here.


Let me ponder this for a bit.


Thanks,

Andres


-- 
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs