nanog mailing list archives

Re: Unusually High traffic from Akamai/Oracle - public-yum.oracle.com


From: Colin Johnston <colinj () gt86car org uk>
Date: Thu, 10 May 2018 08:39:08 +0100

latest yum oracle linux applied ok today fine, 153mb
Have not seen looking content either

Colin


On 9 May 2018, at 20:48, James Stahr <stahr () mailbag com> wrote:



Hi,

Since I'm not a customer of either organization, I'm reaching out to NANOG for a contact and perhaps others may also 
be experiencing similar symptoms over the past 3-4 weeks.  The situation appears to be that customers of ours have 
Oracle Linux and when they attempt to download updates, their traffic goes through the roof for hours on end.  While 
researching this phenomenon, I found this discussion which coincides with the traffic I've seen, however there is no 
mention of excessive traffic resulting from this "corruption" nor have their been any additional reports:

https://community.oracle.com/thread/4138810


Currently, I have two customer environments which are hitting about ~2Gb/s when normally their traffic levels are 
nearly zero.   At first I thought it was an isolated incident but then we observed the same issue with another 
customer.  All of this traffic is coming from 23.35.204.188:80, which belongs to Akamai.  Since that's somewhat of a 
dead end, we examined the hosts which are requesting the data from Akamai and found that they are all Oracle Linux 
boxes and it's a yum process on Oracle Linux which appears to be repeatedly downloading the same content for hours on 
end:


[root@xyzzy noc]# netstat -plutan | grep :80
tcp        0      0 172.16.122.112:14272        23.35.204.188:80            ESTABLISHED 58880/python
[root@xyzzy noc]# ps auxww | grep python
root     41015  0.0  0.3 401940 52044 ?        S    Apr30   0:02 /usr/bin/python2 
/usr/share/system-config-lvm/system-config-lvm.py
root     58880 59.7  1.0 479680 164140 ?       R    18:24  27:18 /usr/bin/python 
/usr/share/PackageKit/helpers/yum/yumBackend.py get-updates none

I can only assume that the data being downloaded is corrupt as this multiple hour download does not consume any disk 
space and because the file(s) are repeatedly downloaded, the logic behind the yum routines are also at fault for 1TB 
of

I don't expect anyone at Akamai to reach out to me since they are simply the middle man here, but I'm hoping that 
someone at Oracle will because the cost to Oracle for Akamai to deliver this junk traffic is not zero and I have a 
hard time seeing how this issue is isolated to our network.  I'd also be interested to hear from anyone else who has 
been seeing traffic spikes from public-yum.oracle.com.


-James Stahr


Current thread: