It is NOT down to internet speed

[mpsamuels]

I would yep.

Okay, I'm sure you probably know some of this Neil (and it is very boring), but I'll go into it anyway for the 3 people in the world that might be interested .

Sadly, but I'm one of those three people!

I haven't had much experience with LBs so where are you suggesting they sit and do their work? By dictating which IP is returned during the DNS lookup is my guess? or am I wrong?

Assuming it is during the DNS request phase and by offering a round-robin approach to which of the three IPs is returned for each lookup, I tried to bypass the sticky session flaw you described by manually editing the hosts file. Although something may have been changed at the back-end which wasn't apparent to me, my experience was that I still ended up presented with a blank page/no response from see's servers.

This was all while running on 10mbps up/150mbps down Virgin Fibre. I didn't get a sniff of a booking page all morning so fail to see how ISP speed has anything to do with it.

[siblin]

Sadly, but I'm one of those three people!

I haven't had much experience with LBs so where are you suggesting they sit and do their work? By dictating which IP is returned during the DNS lookup is my guess? or am I wrong?

Assuming it is during the DNS request phase and by offering a round-robin approach to which of the three IPs is returned for each lookup, I tried to bypass the sticky session flaw you described by manually editing the hosts file. Although something may have been changed at the back-end which wasn't apparent to me, my experience was that I still ended up presented with a blank page/no response from see's servers.

This was all while running on 10mbps up/150mbps down Virgin Fibre. I didn't get a sniff of a booking page all morning so fail to see how ISP speed has anything to do with it.

Actually not quite - i think the 3 DNS entries correspond to 3 live loadbalancers (usually these would be backed up by a hot standby). DNS can be used to do a crude load balance but that assumes every operating system either picks the first host retuned on the list, or one at random (which isn't always the case). DNS requests are also cached by the client according to a time-to-live value, and i know up until recently windows did some nasty things (like caching bad DNS entries and ignoring the TTL value and caching for 24 hours).

LB's are more like complicated firewalls. If you know what NAT is, or you might have come across port forwarding on your home router, the concept is pretty similar. A request comes in to the load balancer and it usually does a port forward (it can work with a next-hop mac address, but this is getting a bit to deep into it and the affect is the same), however unlike a home broadband router, or normal DC firewall, it uses the load balancing logic to determine where the traffic is forwarded to.

How see have got their LB's configured, and what credentials they use to make a session sticky i'm not sure. It may be based on all the session criteria including source port, in which case anything like restarting a browser (or even a full refresh on some browsers may change the logic). I hope its not as crude as source IP address, as this would get you stuck without much hope unless you have access to a 2nd connection. This may be the case given some peoples reports of 3G success.

[eFestivals]

I hope its not as crude as source IP address, as this would get you stuck without much hope unless you have access to a 2nd connection. This may be the case given some peoples reports of 3G success.

From what people have posted, it seems to be the case that there's a 20 minute time-out happening - cos lots of people seem to have reported having a white screen for ages till around 9.23, when they started getting thru. (They perhaps had a working 'queue' screen for the first few minutes).

It might be a time-out within the browser design, but I don't think so. The number of mentions of 9.23 seems to be quite high (far higher than the mentions of any other single time in the threads I've read), and that suggests to me that something at See's end wasn't properly operational until that time .... and then they all suddenly got thru, most got tickets, and then it was very quickly sold out.

[siblin]

From what people have posted, it seems to be the case that there's a 20 minute time-out happening - cos lots of people seem to have reported having a white screen for ages till around 9.23, when they started getting thru. (They perhaps had a working 'queue' screen for the first few minutes).

It might be a time-out within the browser design, but I don't think so. The number of mentions of 9.23 seems to be quite high (far higher than the mentions of any other single time in the threads I've read), and that suggests to me that something at See's end wasn't properly operational until that time .... and then they all suddenly got thru, most got tickets, and then it was very quickly sold out.

Sounds reasonable. I've come across 1200 seconds as a TCP connection timeout on high-load servers before too. It could be as simple as the servers taking that time to flush the table and freeing up some sessions.

[mpsamuels]

Actually not quite - i think the 3 DNS entries correspond to 3 live loadbalancers (usually these would be backed up by a hot standby). DNS can be used to do a crude load balance but that assumes every operating system either picks the first host retuned on the list, or one at random (which isn't always the case).

Understood but surely my attempt at releasing my machine from the stuck session by approaching one of the other load balancers through manually adjusting my DNS should have helped get me a page rather than just end up with another stuck session, no?

[siblin]

Understood but surely my attempt at releasing my machine from the stuck session by approaching one of the other load balancers through manually adjusting my DNS should have helped get me a page rather than just end up with another stuck session, no?

Yep. Though I think i ended up with three duff connections this last time around! I think there are usually many more dead servers than working ones.

[guypjfreak]

look i think us oldens should have an extra day to get our tickets because it takes US ages to type things in to the forms .......its ok for you youngsters your fingers move like lightning but ours dont so please can us over 40s have our Tday a day before Tday ....thanks ah