ghosting, cachepilot, and switches

[PEO]

Hi gang, OK here is a few questions I am searching answers for.

 

during the Easter holidays we decommissioned our old servers (4 year old) and have commissioned the following

 

SAN 4TB Hitachi RAID 6

4 Intel servers

apple Xserver 2TB Mirror

3COM 5000 series switch

UPS cant remember the model

11 tape backup solution which is configured to backup daily, weekly, monthly, annual (on hold) and ghost images (on hold)

 

all is sitting in a nice 42U rack

 

each server is using ISCSI on VLAN and the Ethernet cards have been teamed and VLAN'ed

 

so performance is really good.

 

My next step is to replace all the 3com switches in the school and update to the latest range.

 

at the moment the servers are connected to the network via 1gb copper.

 

As I am happy with 3com I will be sticking with them. Is anyone using the 4000 series which you can use stacking, I'm also thinking about aggregating the servers 1gb connection to the the network so we have 2gb link. I haven't looked at switches in some time so I feel a little out of my depth at present in what to go for and how to configure. I must note we already have a fiber backbone in place.

 

 

 

Ghost imaging

 

I noticed while we were ghosting it was taking 5 hours to send a an image to a room of 30 PC's (14GB 30mb/m) I managed to track down the problem. all traffic was going through the cache pilot which is set as out gateway 10.64.64.4. so I disconnected the switch from the network and we suddenly started to see ridiculous speeds again.

 

Has anyone else come across this problem? could the gateway be set differently so it points to our router? I'm confused and worried as soon as we reconnect the cache pilot it has a noticeable effect on the network.

 

so far the new servers have been performing really well without any error messages. the weakest link now is our switches. Myself and the IT coordinator have talked about this and know that this financial year we need to replace the switches.

[matt40k]

No idea about the 3com stuff, only person I know who ever really done thing interesting (ie managed switches) with them replaced them all with Cisco.

 

As for the cachebox thing, only think I could thing of is somebody set the router\core switch to route all the traffic (going outside of there subnet) though the cachebox. I idea is to force everyone to use the proxy server rather then bypassing it. I've seen it a few times, but to be honest I can't really see how that would work in your case and it's generally considered bad practice to do. Funny how it still works (and better) once you remove it.

[SYNACK]

Agregation for the servers is a really good idea as it increases not only speed but reliability. The 3Com gear is capible and even though I have not dealt with it directly you should still be able to build out a nice fast system with them. Sticking with them will also mean noo need to rebuy your fibre modules.

 

The gateway issue is interesting, how is your internal network structured, do you have multiple subnets and can you do a tracert on one of the imageing computers to the server to see where the packets are going. Also are you using multicast or unicast for the imaging?

[PEO]

No idea about the 3com stuff, only person I know who ever really done thing interesting (ie managed switches) with them replaced them all with Cisco.

 

As for the cachebox thing, only think I could thing of is somebody set the router\core switch to route all the traffic (going outside of there subnet) though the cachebox. I idea is to force everyone to use the proxy server rather then bypassing it. I've seen it a few times, but to be honest I can't really see how that would work in your case and it's generally considered bad practice to do. Funny how it still works (and better) once you remove it.

 

if the cachepilot is disabled in IE you get an error page, only way to get out on to the net is if you set a static IP address and set the gateway to point to the router. so no user can not get out on the net.

 

All out 3com switches are managed.

[PEO]

Agregation for the servers is a really good idea as it increases not only speed but reliability. The 3Com gear is capible and even though I have not dealt with it directly you should still be able to build out a nice fast system with them. Sticking with them will also mean noo need to rebuy your fibre modules.

 

The gateway issue is interesting, how is your internal network structured, do you have multiple subnets and can you do a tracert on one of the imageing computers to the server to see where the packets are going. Also are you using multicast or unicast for the imaging?

 

are you saying If I ghost a machine and tracert that machines IP address from machine I can see where the trafffic leeping around? all machines, switches etc is on the same subnet 255.255.240.0

 

agregation is def the way forward, at the moment I want to aggregate the racks new 3com switch to a dell switch, is this possible with mixed brands? I've heard that this can be a little tricky

[matt40k]

are you saying If I ghost a machine and tracert that machines IP address from machine I can see where the trafffic leeping around? all machines, switches etc is on the same subnet 255.255.240.0

 

agregation is def the way forward, at the moment I want to aggregate the racks new 3com switch to a dell switch, is this possible with mixed brands? I've heard that this can be a little tricky

 

What IP is the server? What is the IP of the client? Is the third block (number) different?

[SYNACK]

are you saying If I ghost a machine and tracert that machines IP address from machine I can see where the trafffic leeping around? all machines, switches etc is on the same subnet 255.255.240.0

 

Yes a tracert from an imaging machine to the server or back the other way while the cach piolet is plugged in whould tell you there the packets are bouncing around on their way (only routers not switches but it may find something). If they are on the same subnet then they should just communicate directly. This should be the same with multicast as at its core it is just a fancy layer2 broadcast that should make it to all stations. I would be investigating the network settings on the imaging computers to make sure they are not attempting anything tricky.

 

agregation is def the way forward, at the moment I want to aggregate the racks new 3com switch to a dell switch, is this possible with mixed brands? I've heard that this can be a little tricky

 

You should be able to do this but you are right that it can be a little tricky. LACP (Link Aggregation Control Protocol) is an accpted standard which is supported by most vendors so it should be possible but it can get messy with passive and active settings for interfaces so it can take a while to get up and running.

[PEO]

What IP is the server? What is the IP of the client? Is the third block (number) different?

 

servers are on 10.64.64.7 etc

 

clients are on 10.64.67. etc etc

 

switch 10.64.65. etc etc

[SYNACK]

Looks like they are all on the same subnet but if the server or hosts still have the wrong subnet mask it may try to route these hence hitting the default gateway.

[sted]

would it be worth removing the default gateway from the ghost servers ip configuration?

[PEO]

good idea will give that a bash

[BaccyNet]

We used to have a similar problem here. It solved itself though when we upgraded all our switches to Procurves and set the multicast filtering option on the computers VLAN.Our cachepilot was clever enough to not touch the packets but the router with the 10mb card seemed to like reading each packet of a ghost image (one room of 10 computers, 10gb image took nearly a full day!!)