Hey everyone,

Thought I would pass my curiosity in a forum as it seems the people I may be dealing with don't seem to understand stuff (I believe).

This place I work likes to host some of their project sites on US hosting and they selected a reasonable sized provider for a few Virtual Servers to which they have placed a project I was in charge of on.

The issue we have been having since it has been on there is the server stops responding at random occasions for 5 mins upwards then returns like nothing has happened.

They say that there is no recorded downtime on their records (do not know where they are monitoring from).

I decided to investigate a bit with some ping plotting and traceroutes on other networks in the country, what I have found is the hop before it hits the server (I assume a router at the DC) appears to change and then it stops responding or lags shockingly for us and other people when this changes.

They are trying to pull the wool over our eyes and say that it is a region specific issue which I don't believe since we are all using National ISPs.

Do you believe like me that it is something on the host's end causing it?

Hi,

We have a script we use for this type of monitoring called sys-snap.

It is a simple bash script that runs in the background and takes a snapshot every sixty seconds of certain system details such as ps, netstat, and vmstat. It then logs all of this to a directory called system-snapshot and rotates the logs every 60 minutes.

You can then check the current log if the server crashes again to get a view of the state of the system just before the crash. Since sys-snap is a bash script, it is easy to customize. If you do not find sys-snap to your liking, you may want to install sysstat instead.

To download and run sys-snap:
wget http://sys-snap.techfiles.us/
chmod +x sys-snap.sh
nohup sh sys-snap.sh &

The logfiles will atleast help you determine if it is a problem with your server or a problem with the router, by giving you log files with all load and process info.

See how you go.

Regards,

Dale

Set up a few 24x7 pings or traceroutes (whatever is easier for you), say 30 seconds apart, ON one of the hosted virtual servers - logging out to few files. Set them up pinging each hop going out from the virtual server to two networks like sprint and mci. You'll soon know if the failure is on the internal network, or after it hits transit networks.

It doesn't let them off the hook either way, but that should be your first test if you 're looking to put your observations through a thorough test.