Should PN upgrade to a new forum ??

[#1]

The problem is people having waited several minutes click Post again.

[#2]

Yeah.  They click post, the response saying "yes, I got that" is dropped by their firewalls because it's coming back from the wrong server, so the browser sits there waiting.  The customer then clicks post again so you duplicate.

[#3]

@'Townman' - I think you described it well enough ! 
Regarding email using my imap client (Thunderbird), over the space of a few minutes my firewall logs can collect rejected incoming packets from say six to eight different Plusnet email server IP addresses, when I would have thought it should be sticking to one.  I think that the only reason email 'appears' to work is that it is polling so often that lost packets and connectivity problems are masked by the subsequent successful fetches.
It is somewhat annoying that my firewall logs are 99% full of Plusnet rejected packets, when I just want to see the logs of real port scans and other connection attempts.  It is also sad that I have had to specifically add router rules to bypass my otherwise wonderful caching http proxy when accessing this forum.
I have complained about this for months, but as with everything Plusnet it is not a priority ... blah blah blah. 
It's probably my own fault for being a smart-arse and building my own internet gateway, rather than using commercial junk 
I do find it somewhat ironic that my internet service provider has the worst configured network of ANY of the services I access on the web 

[#4]

Quote from: Kelly

Yeah.  They click post, the response saying "yes, I got that" is dropped by their firewalls because it's coming back from the wrong server, so the browser sits there waiting.  The customer then clicks post again so you duplicate.

I can understand that but when posting with an attachment and while the browser is still saying waiting it is actually posts but only after several minutes delay,
Surely if it was only due to the browser receiving a response saying I have got it it should actually post almost immediately and show up in the other browser window which it doesn't

[#5]

Quote from: Oldjim

while the browser is still saying waiting it is actually posts but only after several minutes delay,

Jim,
There does not need to be several minutes delay.
Kevin

[#6]

What Jim said is factually correct - I've seen it on a number of occasions.
You make a post with an attachment and then when it hangs open another browser tab and view and keep refreshing the forum index. It can be several minutes before you see your post appear as the latest post. Meanwhile the original tab is still hung and typically stays hung.

[#7]

If you make an assumption that perhaps the load balancer problem is faulty in both directions (not just on replies), i.e. your browser might be having it's output directed (incorrectly) to more than one community server, then -
What if the browser thread that initiated the post, sits waiting for effectively a session (and file upload) complete signal, but that never happens because of the load balancer problem.  It will just sit there indefinitely because the packet sequence has been lost.
The uploading attachment file will be occurring in a separate thread using a file transfer protocol.  The file transfer protocol will be much more robust than the browsers simple http transfer mechanism, so will be sitting there doing multiple retries of uploading packets (due to repeated failures because of lost data in the load balancer) BUT because of the persistent nature of the retries there is a good probability that the file transfer will eventually complete despite the high packet loss.  This would explain the big delay but likely success.
The other browser tab used for monitoring the original posting browser tab, is merely reporting the progress of the file transfer, delayed by the repeated stalling and retries due to load balancer lost packets.

[#8]

I don't think there's any separate thread or some file transfer protocol - I expect it's just a http post request.

[#9]

Purleigh's last post is IMHO the most plausible explanation so far.   A year or so back a large National company I worked for had similar problems with a large Oracle database with load balancing and a number of Oracle options to ensure (well the intention was there?) that ALL transactions were recorded and no sessions lost.  It took a long time and several attempts to finally identify THE root cause in load balancing.
As is expected once the cause was found, the 'fix' was trivial - but an awful lot of knowledge was gained on the journey.
M.  

[#10]

I would argue that this forum is sufficiently small (in terms of transactions per second) that it probably doesn't need to be behind a load balancer, unless the hardware is really ancient or the server function is being shared with something else.

[#11]

Quote from: purleigh

I would argue that this forum is sufficiently small that it probably doesn't need to be behind a load balancer, unless the hardware is really ancient.

I would agree.  Another consideration will be other services being supported on the same Server farm adding to the confusion?  Without some recent picture of the Network configuration it's all conjecture.  The problem I referred to was only finally resolved with a small team from the Hardware; Oracle; Network software an OS collaborating to solve a common problem.  Perhaps that time is near for Plusnet i their capability is to remain untarnished.
Maurice

[#12]

Yes, my solution for the forum issue does not help resolve the similar problems of the email imap servers responding on half a dozen IP addresses, and which probably still needs to be behind a load balancer in case all 700,000 customers all access their email at the same time.
However, fixing one thing at a time would at least be making some progress, especially as I've not seen anyone else complaining about their firewall logs being full of broken imap requests, probably because each users email client will continue to poll and hence is likely to recover/succeed without reporting connection issues.

[#13]

PN have said in the past that there's a lot of traffic for the blogs (God knows why!).

[#14]

I don't use the Plusnet email servers, although I did point out the spurious packets several months ago. They haven't even fixed the reverse DNS entries yet either.
Balancing the load over the Plusnet customers at least could be done based on which gateway you're connected to (assuming most of the traffic is from Plusnet broadband connections). Last time I checked, there are 2 community servers in PCL and another 2 in PTN. But it's not being done like that. You go to 212.159.8.110 (PTN) or 212.159.9.110 (PCL) at random regardless of which gateway you're on.
Or alternatively, they could just have community.plus.net resolve to more IP addresses, and rely on "DNS round robin" to spread the load. It's already partly done like that since community.plus.net resolves to both 212.159.8.110 and 212.159.9.110.

[#15]

Looking at the tracerts the servers look they they are in the Sheffield data centres, not London.