Analyzing dropped packets

[Aviv Greenberg]

On Wed, Mar 11, 2009 at 13:49, Michael Green <mishagreen at gmail.com> wrote:
> On Wed, Mar 11, 2009 at 11:36 AM, Aviv Greenberg <avivgnet at gmail.com> wrote:
>> What is the server used for? What traffic is comming into the server
>> (small packets? full MTU?)? at what rate? How many RX descriptors is
>> your driver set for? Do you have 802.3X flow control enabled (both
>> ends)? Do you see any abnormal CPU usage when the packet drops occur?
>>
>
> I'm not quite sure how to check out all these things...

Mainly using ethtool - read the man page

>
> The server is doing backups. Naturally most of the traffic happens
> during night hours.
>
>
>> But more fundementally, why are you even concerned with 2916/689587830
>> packet drops? On most scenarious, this is pretty normal behaviour.
>>
> I'm concerned because occasionally the backup application shuts its
> TCP port on which it listens. It seems to happen after a considerable
> amount of drops/errors have accumulated on the interface. The only way
> out then is to recycle the app. I suspect that these errors might be
> the indication of the problem that causes abnormal application
> behavior.
>
>
>> i would be more curious why there are 2900 errors :)
>>
>
> I realize there might be a million different reasons for these errors.
> How can I investigate them?

1. Use ethtool to set and query device settings
2. Use ethtool -S and netstat -s to get error statistics
3. Use dmesg to see if there were any errors printed by the device
(hangs, watchdog messages, etc)

>
> _______________________________________________
> Linux-il mailing list
> Linux-il at cs.huji.ac.il
> http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
>



-- 

Bob Hope  - "Middle age is when your age starts to show around your middle."