Probable packet loss - how to diagnose

Hello it seems my provider has trouble with my ISP since last week, my web sites are "hanging" somehow (pages not completely loaded, slowliness, etc.).

Needless to say my web server is on good health (Load average below 0.5 all the way, and surfing is very fast on local network)

I've made so diagnoses myself on the situation to arise a packet loss situation :

###########################################
Here's a Trace route from my routeur to my host through my ISP (Free - France), are the stars *  packet lossesz ?

Trace route to 193.19.216.145 over a maximum of 30 hops:
1  6.894ms   *   0.008ms   81.57.32.254
2  1.865ms   32.232ms   *   213.228.5.254
3  *   *   0.048ms   212.27.51.82
4  0.213ms   0.295ms   23.664ms   212.27.50.37
5  0.323ms   0.260ms   *   212.27.50.34
6  0.790ms   0.376ms   0.559ms   213.228.15.67
7  0.426ms   0.622ms   1.025ms   193.19.216.145
Trace route completed.

###########################################
I've also made a path ping to my remote server on a windows client and it appears that there is also packet loss on the 8th node :
ovanet.freeix.net

Détermination de l'itinéraire vers 85.12.147.20 avec un maximum de 30 sauts.

  0  WORKSTATION [192.168.1.4]
  1  Administrator [81.57.32.137]
  2  laborde-1-81-57-32-254.fbx.proxad.net [81.57.32.254]
  3     *     p19-6k-2-a5.routers.proxad.net [213.228.5.254]
  4  vlq-6k-2-v810.intf.routers.proxad.net [212.27.51.82]
  5     *     th1-6k-2-v804.intf.routers.proxad.net [212.27.50.37]
  6  cbv-6k-2-v802.intf.routers.proxad.net [212.27.50.34]
  7  ldc-6k-1-a0.routers.proxad.net [213.228.15.67]
  8  ovanet.freeix.net [213.228.3.195]
  9  ge-ve23-2-1.ar1.rdb.fr.eurowan.net [193.19.216.146]
 10  85.12.147.20

Traitement des statistiques pendant 250 secondes...
            Source vers ici  Ce noud/lien
Saut RTT    Lost/Sent =  Lost/Sent  =  Adresse
  0                                           WORKSTATION [192.168.1.4]
                                0/ 100 =  0%   |
  1  ---     100/ 100 =100%   100/ 100 =100%  Administrator [81.57.32.137]
                                0/ 100 =  0%   |
  2    6ms     0/ 100 =  0%     0/ 100 =  0%  laborde-1-81-57-32-254.fbx.proxad.net [81.57.32.254]
                                0/ 100 =  0%   |
  3    7ms     0/ 100 =  0%     0/ 100 =  0%  p19-6k-2-a5.routers.proxad.net [213.228.5.254]
                                0/ 100 =  0%   |
  4    8ms     0/ 100 =  0%     0/ 100 =  0%  vlq-6k-2-v810.intf.routers.proxad.net [212.27.51.82]
                                0/ 100 =  0%   |
  5   10ms     0/ 100 =  0%     0/ 100 =  0%  th1-6k-2-v804.intf.routers.proxad.net [212.27.50.37]
                                0/ 100 =  0%   |
  6    8ms     0/ 100 =  0%     0/ 100 =  0%  cbv-6k-2-v802.intf.routers.proxad.net [212.27.50.34]
                                0/ 100 =  0%   |
  7    7ms     0/ 100 =  0%     0/ 100 =  0%  ldc-6k-1-a0.routers.proxad.net [213.228.15.67]
                                0/ 100 =  0%   |
  8  ---     100/ 100 =100%   100/ 100 =100%  ovanet.freeix.net [213.228.3.195]
                                0/ 100 =  0%   |
  9    7ms     0/ 100 =  0%     0/ 100 =  0%  ge-ve23-2-1.ar1.rdb.fr.eurowan.net [193.19.216.146]
                                0/ 100 =  0%   |
 10    7ms     0/ 100 =  0%     0/ 100 =  0%  85.12.147.20

Itinéraire déterminé.

###########################################
From the Host trace route tools it appears some trouble, you can test by yourselves, what mean the losses on their stats ?

http://www.eurowan.net/traceroute-mtr.cgi?t=213.228.0.42

I'm no routing expert and my interpretations may be wrong, what do you think about this situation, are those statistics proofing any trouble I should ask my host about ? Do you have any hints ?

Thanks
FFTAsked:
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

pablouruguayCommented:
no. maybe can we a hidden routes in the routes... but many times are congestions... :(

but the pings are ok 7ms is a good time but you have problems with this 2 servers

cbv-6k-1.routers.proxad.net              50%     3    6    0.32    0.36    0.43
th1-6k-2-v802.intf.routers.proxad.net    17%     5    6    0.58    0.87    1.79

try to contact your netadmin in your ISP and explain your problem... many times the netadmin didnt realize about the problems of a one route.

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
harbor235Commented:
There are several reasons why you will sometimes see * inyour traceroutes.

1> Are you using microsoft tracert or *nix traceroute?
MS actually uses ICMP for their tracerts and routers will throttle ICMP messages if the are busy
*nix uses traceroute, and tracerotue uses udp.

2> Some providers turn off/filter certain ICMP messages type

I do see loss from the traceroute url provided, however, this may or may not be valid if the sending host is a microsoft platform Ping was not working on www.eurowan.net. Ideally you need to perfrom extended pings on each hop from your source to each destination in the path, try 1000 packets for each hop. Thsi will give you a better feel for packet loss, this form of ICMP will work because all ISP allow echo request/replies

harbor235 ;}
verivoxCommented:
Hi...

if you want to see how reliable your ISP is, try setting up a "smokeping".
You can see an example at http://smokeping.dirtylittleserver.net 

Smokeping is a really fine way to see what's up on your network.

Cheers,
Sebastian
C++ 11 Fundamentals

This course will introduce you to C++ 11 and teach you about syntax fundamentals.

FFTAuthor Commented:
Thanks everyone for all these tips, after some research it appears that icmp ping is not a proof of packet loss since many isp limit their router about icmp traffic, i've made up a simple page :
http://www.homestudio.fr/debug_simple.php

I just wanted to know if for you the simple fact to click BACK OR FORWARD made some pages hanging for you (easier to see with firefox as IE must wait all the code to de downloaded to display the page) ?

I tried with 4 different ISP's (Free - Wanadoo - Club Internet - Neuf Telecoms) with the same results, so there should be some problem between the host and these ISP's (my host did not replied so far).

Some tests i've made :
- Copied a big web site to another server on the same LAN to see if it is an hardware/software problem : no, I also get unexplained latencies while surfing on different configuration.
- Use different version of Apache/Php (2.053/2.2 - php 4.4 to 4.4.2, etc.) -> same results
- Improved mysql configuration a bit :
  key_buffer              = 256M (was 16M before)
  table_cache             = 256 (was 64 before)
- Copied the big web site to another server on another location : much better... so i really tend to believe it's my host here...

I've seen some slights changes while modifying the mysql conf file but nothing spectacular so far...

I've made some UPD (can't do TCP pings with WXP/SP2) pings though these different ISPs, some have great losses (Free..), some not (Club-internet), what I do not understand is : when I download a big file (100 mo) from my server it goes fine, but with a simple 30Ko dynamic php/mysql page I got hangs all the way.

What is your opinion ?

Thanks
FFTAuthor Commented:
Can someone give me reply on this ?

It clearly appears that the php/mysql process is minimal, thus it should be a data transmitting problem on the network.

I've made different copies of the same test php file on two different networks and there are great difference in term of response times..

To do the test simply load the pages and click several times on the "FORWARD" link on the upper right to make some php/mysql calls, on the OVH network the responses is very fast, on OVANET it is sluggish and the page hangs at the middle for no reason... (not all the time but often)

All the servers are similar in term of hardware/software and bandwith and nothing can justify at my level such differences (absolutely  NO overload), I just can says that during some periods of the day things are getting better for no reason but it never last long (30mn...)

The test is reproductible on 4 different French ISPs with no packet loss problem (UDP/TCP)

My Host (Ovanet) says it doesn't find any problem on his network...

I'm just getting mad with this stuff please help...

OVANET
85.12.147.20 - Server #1
http://www.homestudio.fr/debug_simple.php
85.12.147.27 - Server #2
http://proxy.maxserv.net/debug_simple.php

OVH
213.186.34.46 -- Server #2
http://homestudio.team-project.net/debug_simple.php
Login : ovanet / password : ovanet

Thanks for your help.
FFTAuthor Commented:
Thanks for everyone help, My provider found a defective cable on his system that was the cause of this problem, It was difficult to diagnose because somehow the route taken was not only through this cable, that's why we had so much "hanging" pages. Anyway no tools really showed up the problem at my level, I've even used pingplotter to send some packet (TCP/UPD not ICMP) from an internal windows 2003 server but the packet were not lost, the traffic was only slowed down...
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Linux Networking

From novice to tech pro — start learning today.