Randomly, CT log server response takes more than 60s

[David Fernandez]

Hi all,

After installing and configuring the sources from GitHub, I've noticed some strange behaviour after monitoring it.

Sometimes requests (just get-sth) responses last more than 60s....while the rest are served in milliseconds. Usually clients time-out is shorter than this, but if not, I see that in this cases, responses took just 60s and some milliseconds.

Looking to the web server logs, I can also see that this errors take place once every 5 minutes (more or less):

(70007)The timeout specified has expired: [client ip:port] AH01110: error reading response.

And if any request arrives at that very moment, the answer  tooks 60s.

And all of this regardless the amount of requests (the same behaviour  with 500 req/s or only monitoring)

I would like to know if this could be related to some of the flags of the ct server, mainly with:

• -connection_pool_max_unused_age_seconds
  
• -url_fetcher_max_conn_per_host_port
  
• -num_http_server_threads
  

Regards,

[Pierre Phaneuf]

Is there any strange log message on the server side when that happens?

Maybe try running with --v=1 for extra verbosity, but that will make
the analysis more difficult...

> --
> You received this message because you are subscribed to the Google Groups
> "certificate-transparency" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to certificate-transp...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.