strange website access problems

[luap.h]

I help admin a website for Mazda 323fs (www.club323f.com)

We put quite a drain on our shared server, so they moved us to a quieter one. Ever since then our members have had temperamental access problems (site not found). At first I though it was because the DNS servers hadn't updated fully yet after the move to a new server, but over a week later it's still going on.

I've got them to try different browsers, clear caches etc, but still happens. There's one guy, when it works, tracert and ping resolve fine, but when the site is 'down' the destination is unknown. But, the guy can log into his work machine from home, and access the site ok - so that infers the site isn't down at all.

There's another guy who can access the site when he uses wireless, but not if he uses wired connections?!

Another says that he can access the site until around 12, and then it disappears?!

The closest thing I can find on google is MTU not set to 1400 - but it seems strange that it would be causing trouble for so many users.

Any one out there have any ideas?

Thanks for any help!!

[alsenior]

give the hosting company a talking to it seams to be there problem

[luap.h]

I did - all I got was

I moved this account onto the fastest and clearest server we have, as it is such a huge pull on
the server that it is on, dedi 2 has had no reported problems over the long weekend and is working fine
at the time of answering this ticket

Dedi 2 85.13.239.15 server load average: 0.20 0.41 0.42

its not our server for sure and I checked the bandwidth and that seems fine too

[Jay]

ask people to trace route to the server during the day then, when they can't access it do the same.

they should be able to see where the problem is. It could be a DNS problem or even a server on route that is overloaded at different times of the day.

[raid1]

If multiple users on different ISP's can't get to it it is likely to be a problem at the hosting company.

You are using ping and tracert to test it, another command worth trying is nslookup

i.e. nslookup www.club323f.com

which will do a dns lookup and might give you more info when it is failing.

[luap.h]

I'll get them to try the DNS lookup thing tonight

It's just really weird that one person can get it at home when they use wireless, but if they connect by ethernet, it doesn't work!!

[Splash]

I thought I replied to this yesterday, and I did... in the wrong frickin' thread...

Alertra tells me

Time (US/Eastern) Checked From Result KBytes Secs Kbps
05/30/2007 05:18:22 London UK OK 7.9 0.27 238.25
05/30/2007 05:18:22 Frankfurt GERMANY OK 7.9 0.45 142.59
05/30/2007 05:18:27 Hong Kong CHINA OK 7.9 5.50 11.73
05/30/2007 05:18:23 Chicago USA OK 7.9 1.71 37.77
05/30/2007 05:18:24 Oklahoma City USA Error N/A 3.03 N/A
05/30/2007 05:18:24 Orlando USA OK 7.9 1.92 33.60

Error when accessing from Oklahoma. Error code is

TIMEOUT 30
DNS www.club323f.com 85.13.239.5 0.407821893692
TCP 80 0.175965070724
ERR Connection to HTTP service failed

LAST_REQUEST = http://www.club323f.com/
STAT_TIME_TO_LAST_BYTE = 0
LAST_MSG = Connection to HTTP service failed
LAST_EXEC_TIME = 2.45029115677
LAST_TIMESTAMP = 1180516849.05
LAST_PROTO_CODE = ERR
LAST_IP = 85.13.239.5
DNS_ANSWERS = 1
LAST_CODE = 3
STAT_TOTAL_BYTES_WRITTEN = 0
DNS_ANSWER_1 = 85.13.239.5
LAST_PROTO_MSG = Unrecognized HTTP error
LAST_HOST = www.club323f.com
STAT_TOTAL_BYTES_READ = 0
TOTAL_RUNTIME = 3.03486990929
STAT_TIME_TO_FIRST_BYTE = 0
LAST_COMMAND = HTTP
TRACEABLE = Y
STAT_CONNECT_TIME = 0
STAT_WRITE_TIME = 0

Not sure what the fault is in all honesty, but I'd push this back to the hosts for investigation...

[luap.h]

I signed up for free trials of Alertra, Alert Site and Host Tracker - and all say that at various times in various countries, the site is not reachable

I got some users to try using a web proxy (Guardster) to see if they could access the site while it was "down" and they could.

My hosting aren't doing too much at the moment

I emailed them the reports from Alert Site and got this

Looking at this issue - from your traceroute you are receiving a timeout at hop 5;

5 29 ms * 30 ms 195.66.226.185

This may cause a page to not load correctly depending on what the actual issue is. This problem relates to a peering issue with BT. We cannot obviously advise of any issues outwith out network though.

Irregardless of whether or not you have a dedicated server if there is in fact a DNS issue this problem will still happen. I have looked within our network and can see no DNS issue with your domain name, or the server.

I would advise getting someone who cannot view the site to use an anonymous proxy server at the time they cannot view it. This will show us whether there is an ISP, Peering, or a DNS issue

Then the next day I told them to put the url into the host check at Host Checker, as I was getting alert mails the site was down

Received responses: 11 OK 3 Fail

if there was a problem with the server it would be 0 OK and 14 fail

it carried on

Received responses: 17 Ok 5 Fail

this is the network and not the server and it could be their end or in between

all I can tell you is its not the server as I said it would be all Fail if it was

I'm not really sure where I can go from here. I'm getting the site users to tell me what ISP they have and if they can get through or not.

[Splash]

In all honesty I'd be taking my site elsewhere...

[Bluecube]

Yup. This does sound like a hosting server problem. If the people you're paying cash to can't sort the problem out then take your custom elsewhere.

[ikonia]

There are many things that can cause this, and to be honest I think your looking in the wrong places.

If it was DNS related it would be perminently failing for some ISP's and not for others, the fact that you've told me that some users can trace router or ping the machine but not get a response through the browser, and looking at the errors that are being reported it seems to suggest that for some reason you webserver is not responding to http requests at certain times.

this could be for any number of reasons, such as.

1.) your using a virtual host on a bigger machine, the statistics show to you are of the resources of your virtual machine - not the overall chassis that could be used by other hosts.

2.) the load you showed me for your machine appears quite high for a web server - its not unreasonable (more so if you are on a virtual host) to suggest that thinks like disc access maybe timing out.

3.) you are running a club or fan site which suggests a forum, which suggests a database which may not be responding in time due to load, poor config so can't display a page

4.) the most obvious - you have a "max http" connection limit set on your webserver which is too low, so even if your server is fine with resources your connections are limited to this setting - any more and it will not display a page.

[Splash]

What Ikonia said basically, but all point to it being a host issue

[luap.h]

hmmm, that's quite interesting

we are on a shared server, I know that much. We do have a forum, and that's why the site was moved in the first place. It was placing such a load on the old server, it ground to a halt and then they kept suspending our account. The hosts then moved us to a quieter server, and ever since we've had these problems. It's not just the forum they can't access however, its the main site too, which is just standard html

The max http thing sounds promising, but you'd like to think the hosting company would know to check this!!

I'll pester them again, see if it makes any difference - thanks for all the help though!

[luap.h]

so under the request of the Host, they asked if a user that had trouble accessing, could get through on a proxy.

Using Guardster they could.

I also got them to supply which ISP they were using

geetee - no problems - Virgin media
parvez - no problems anywhere - Virgin Media
Murray - no problems at two different locations - Virgin Media

NMMotorsport - no problems anymore - Bt
Canthinkof12 - No Problems Ever - BT

maylander739 - problems all the time - Talktalk

ErrNee - Geting better after refreshing - Wa*kadoo / Orange
Mark.T - period problems - Orange BB
andym323 - Orange BBand - same probs as others, slow loading of pics
etc

Adam - Problems, refreshing ~5 times usually gets it going - AOHell
yourcodenameisnick - Problems, refresh a few times - AOL

yourcodenameisnick - A few problems every hour or so - JHCS (local company)

mmpob - no problems ever - Pipex

So I got this reply

looking at the list it seems to be all the big ones are ok, but the talktalk and rubbishrubbishrubbishrubbishadoo / Orange
seem to have problems, this is totally down to them as all the others work fine, i'm sorry to
say you on the nail with just one of those things
Regards,

[luap.h]

ok weird 'new' development

One user commented that on BT everything was fine until they clicked on "show all smilies" in the forum. Then the website was unavailable.

I tried it myself (pipex) and exactly the same thing happened!! No website, no tracert, no ping - nothing!

I rebooted my router, got a new IP address and everything works fine again!!

Tried it again, and same thing happened! Reboot, new IP, its all ok again!
I had a look at how many smilies there are, there are just over 300!! yeow!!
You think maybe when a user clicks on show smilies, 300 odd connection requests fire off, the host goes WTF and then blocks the IP ?