timer_resolution / $request_time
arthur.blake at gmail.com
Fri Aug 27 23:17:19 MSD 2010
that doesn't actually work:
[emerg]: unknown "request_time000" variable
On Fri, Aug 27, 2010 at 2:41 PM, Arthur Blake <arthur.blake at gmail.com>wrote:
> Is there anyway to get the request time in microseconds resolution in order
> to emulate Apache's %D log format option?
> It looks like the only option at this time is to just fake it with
> something like "$request_time000" (since 1 millsecond = 1000 microsecond)
> On Thu, Aug 26, 2010 at 6:59 PM, Maxim Dounin <mdounin at mdounin.ru> wrote:
>> On Thu, Aug 26, 2010 at 11:23:47PM +0200, Xavier Martin wrote:
>> > Hello,
>> > I've been using this settings for a while:
>> > timer_resolution 100ms;
>> > log_format combined_time '$remote_addr - $remote_user [$time_local] '
>> > '"$request" $status $body_bytes_sent '
>> > '"$http_referer" "$http_user_agent"'
>> > ' $request_time';
>> > access_log /var/log/access.log combined_time;
>> > Right now i'm needing more precise timestamp in logs
>> > => finding requests that took less than 50ms to complete.
>> > I would like to know what are the drawbacks of using a lower number
>> > (i.e 10ms, 0ms or not setting at all timer_resolution) and overhead that
>> would cause (if any).
>> With timer_resolution unset (set to 0, default) nginx will call
>> gettimeofday() on every event loop iteration. With
>> timer_resolution set nginx will schedule
>> gettimeofday() calls at specified interval.
>> Obviously changing it from 100ms to 10ms would cause 10 times
>> more gettimeofday() calls. But most likely you won't notice.
>> But actually I would recommend using the default (i.e. unset).
>> It's not really different from 10ms on loaded servers and wouldn't
>> cause extra work on otherwise idle servers.
>> Maxim Dounin
>> nginx mailing list
>> nginx at nginx.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the nginx