Portal Home > Knowledgebase > Articles Database > High Ping + Packet Loss at Dedicated Server
High Ping + Packet Loss at Dedicated Server
| Posted by Urosino, 08-01-2009, 06:48 AM |
| I am facing high ping and packet loss issues with a server hosted at hetzner.de (CentOS + WHM)
For some unknown reason pings go high and stay like this (average 1500ms) until I reboot the server.
Here is a screenshot of tracert from server to me: http://www.imagebam.com/image/37409243863747
Any tips and opinions are more than welcome.
Thank you in advance!
|
| Posted by SoutheastWeb, 08-01-2009, 07:26 AM |
| Just by looking at your traceroute the latency to your first hop is extraordinarily high.
What are your loads on this server? How fast is your physical port speed?
A little minor troubleshooting should point you in the right direction. Try checking your mrtg/rtg graph, or running top and iostat. Do you see a high percent of IOWAIT?
When in doubt open up a trouble ticket with your provider and have them investigate the issue.
|
| Posted by 040Hosting, 08-01-2009, 07:34 AM |
| Those high values solved by a reboot; would make me advise you to contact your server management company and let them look into it; either the server is to busy (to small for to many requests) or you may have some kind of attack to your server. Really there are to many reasons why this can happen.
|
| Posted by Flumps, 08-01-2009, 07:42 AM |
| also, speak to your provider and ask them if any of the other servers in the same cab are bursting to port speed. (common cause of packet loss). sounds like to me they have over sold there network either that or a dogy switch some where.
however considering you said its your server that gets sorted by a reboot. im more inclined to say its a dogy script or somthing on ur box giving you issues..done any scans for virues, trojans etc?
|
| Posted by Urosino, 08-01-2009, 07:51 AM |
| Correct, that's why I believe it's a server-side issue.
Server loads are low (0.30 0.90 1.51), speed eth0: negotiated 100baseTx-FD flow-control, link ok
I have some troubles configuring mrtg, but here are other details.
top: http://www.imagebam.com/image/fa69e443868602
iostat: http://www.imagebam.com/image/bb70cd43868656
I have, but they confirm it's nothing wrong with their network or HW, because after rebooting works normally.
Any suggestions on what to do next?
|
| Posted by 040Hosting, 08-01-2009, 08:14 AM |
| Did you check your connections to your server ?
i.e.
might want to replace : grep tcp with : grep "tcp\|udp"
|
| Posted by Urosino, 08-01-2009, 08:23 AM |
| Does it look "ok" to you?
udp: http://www.imagebam.com/image/50051043871198
tcp: http://www.imagebam.com/image/cd545a43871266
|
| Posted by 040Hosting, 08-01-2009, 08:35 AM |
| Seems to be okay as well; one other thing you may have a problem with is the firewall; if you installed any; did you try what happens when disabling it; or completely flushing all added rules ?
|
| Posted by Urosino, 08-01-2009, 08:36 AM |
| I haven't installed any additional firewall. Server was built a week ago, and working fine till yesterday.
|
| Posted by 040Hosting, 08-01-2009, 08:39 AM |
| I would install a firewall asap to block any unwanted traffic. I would suggest to look at CSF; http://www.configserver.com/cp/csf.html
and let someone look at your server with server management skills to find out where this comes from.
|
| Posted by Urosino, 08-01-2009, 08:57 AM |
| Do you recommend any management service I should contact?
|
| Posted by 040Hosting, 08-01-2009, 09:12 AM |
| There are plenty of companies; try to look at the WHT wiki for managed companies; and you will see a whole list of companies you can contact.
|
| Posted by SoutheastWeb, 08-01-2009, 07:04 PM |
| The load seems fine, same with the iostat output. Is the server still showing high latency? Do these captures again when the server is slow. Your provider should have a traffic graph, if not try making a request. Are you bursting higher than the rate limit or physical port speed? There are any number of reasons for high latency, keep digging.
|
| Posted by Scott.Mc, 08-01-2009, 07:40 PM |
| Provide the output of lspci, I believe hetzner use realtek nics which have an issue with the r8169 driver provided by centOS. Using the old version r8168 will correct it.
|
| Posted by Urosino, 08-02-2009, 04:56 AM |
| I have checked a graph, nothing special showing, no spikes.
Here is output of lscpi: http://www.imagebam.com/image/6b652e43965908
Does this make sense Scott?
|
| Posted by Scott.Mc, 08-02-2009, 07:27 AM |
| So it is the realtek driver, try what I said and install the r8168 driver which should correct the issue.
|
| Posted by SoutheastWeb, 08-02-2009, 07:46 AM |
| Looks like a Core i7, your nic should compatible with CentOS (RT811/8168B).
Paste the output of netstat -i, ifconfig eth0, and don't forget to cover up your IP address.
|
| Posted by Scott.Mc, 08-02-2009, 08:06 AM |
| I found a link for you,
http://wiki.hetzner.de/index.php/Ins...der_CentOS_5.3
|
| Posted by Urosino, 08-02-2009, 12:13 PM |
| I will try to install driver and get back to you.
Just a question Scott, do I need to follow "Using kmod-r8168 from elrepo.org on CentOS" step only or "Installation from source" as well?
Thank you Scott and SoutheastWeb.
|
| Posted by Scott.Mc, 08-02-2009, 05:15 PM |
| You can use either, no need to both. The kmod-r8168 will most likely be a prebuilt module for your CE kernel so you will only need to install this, this is likely the easier option for you.
The other option is to install from source which base in mind you will need the kernel rpm to provide the sources in /usr/src before you can compile against it.
|
Add to Favourites 
Print this Article
Also Read
