Cache hit gives lower result than cache miss

[Noam Schiff <no...@gmail.com>]

Hi list,

I'm seeing a very weird behavior (at least to me it looks weird) while
running thousands of requests/s through the Traffic Server machine.
When configuring the Traffic Server not to use cache I'm seeing around 30k
of requests/s. While when I configured cache I see only 24k (when I'm
trying to increase the load I suffer from packet loss).

Did anyone see such behavior? Is this a configuration issue?

Any advise,
-Noam

Re: Apply configuration changes

[Leif Hedstrom <zw...@apache.org>]

On Sep 3, 2013, at 8:30 AM, Eddie <zt...@yahoo.com> wrote:

> Hi Guys,
> 
> I have couple of questions and hope someone can help ...
> 
> 
> 1: Can I do "configtest" before I apply configuration change made to remap.config on a running ATS instance? 


Unfortunately not, this is a feature we really need.

> 2: Do I have to restart ATS for "map" rule change? So far, seems to me I have to do it. Otherwise, the new rule added is not picked up.

You shouldn't have to. After editing or pushing the new remap.config, all you should have to do is traffic_line -x (running as a user with appropriate permissions).

> 
> The problem I am facing is that if the new "map" rule is not created correctly,  the restart will fail and cause outage. But there is no way
> for me to test the configuration.

If I recall, traffic_line -x will not replace the old configuration with a broken one. Hopefully it would also say something in the syslog and/or error logs. If you do a restart, that is obviously not an option, it'll simply refuse to restart with a broken configuration.

-- Leif

Re: Apply configuration changes

[Eddie <zt...@yahoo.com>]

Thanks for the reply ... i am thinking about setting up a staging server.


________________________________
 From: Alan M. Carroll <am...@network-geographics.com>
To: Eddie <us...@trafficserver.apache.org> 
Sent: Wednesday, September 4, 2013 7:30 PM
Subject: Re: Apply configuration changes
 

I don't know of an official way to test the config. But you could just install ATS on some other test box and run it with the config to verify the validity. You don't even need to two network cards or anything else.

> The problem I am facing is that if the new "map" rule is not created correctly,  the restart will fail and cause outage. But there is no way for me to test the configuration.
> Any idea how I can do it?

Re: Apply configuration changes

[Eddie <zt...@yahoo.com>]

Thanks for the reply ... i am thinking about setting up a staging server.


________________________________
 From: Alan M. Carroll <am...@network-geographics.com>
To: Eddie <us...@trafficserver.apache.org> 
Sent: Wednesday, September 4, 2013 7:30 PM
Subject: Re: Apply configuration changes
 

I don't know of an official way to test the config. But you could just install ATS on some other test box and run it with the config to verify the validity. You don't even need to two network cards or anything else.

> The problem I am facing is that if the new "map" rule is not created correctly,  the restart will fail and cause outage. But there is no way for me to test the configuration.
> Any idea how I can do it?

Re: Apply configuration changes

["Alan M. Carroll" <am...@network-geographics.com>]

I don't know of an official way to test the config. But you could just install ATS on some other test box and run it with the config to verify the validity. You don't even need to two network cards or anything else.

> The problem I am facing is that if the new "map" rule is not created correctly,  the restart will fail and cause outage. But there is no way for me to test the configuration.
> Any idea how I can do it?

Apply configuration changes

[Eddie <zt...@yahoo.com>]

Hi Guys,

I have couple of questions and hope someone can help ...


1: Can I do "configtest" before I apply configuration change made to remap.config on a running ATS instance? 
2: Do I have to restart ATS for "map" rule change? So far, seems to me I have to do it. Otherwise, the new rule added is not picked up.

The problem I am facing is that if the new "map" rule is not created correctly,  the restart will fail and cause outage. But there is no way
for me to test the configuration.

Any idea how I can do it?

Thanks

Eddie

Re: Cache hit gives lower result than cache miss

[Yongming Zhao <mi...@gmail.com>]

unless you are hitting some cutting edge situation, you should not get that result, how is your testing?
 is that request all cached on disk? how about the disk io utils%? how about your request hit ratio?


在 2013-8-14，下午9:38，Noam Schiff <no...@gmail.com> 写道：

> Hi list,
> 
> I'm seeing a very weird behavior (at least to me it looks weird) while running thousands of requests/s through the Traffic Server machine.
> When configuring the Traffic Server not to use cache I'm seeing around 30k of requests/s. While when I configured cache I see only 24k (when I'm trying to increase the load I suffer from packet loss).
> 
> Did anyone see such behavior? Is this a configuration issue?
> 
> Any advise,
> -Noam