Rehosting Work on the Forum Virtual Server

[TerryE]

By way of background, I am the very aged IT geek that has done all of the SysAdmin since we first set up BuildHub about 9 years ago.  I am no longer involved in Forum Management or politics; I just in keep the forum up and smoothly running, pro bono. i have also administered and run a number of community forums and wikis over the last few decades, mostly using open source forum packages such as phpBB and MediaWiki.

 

However when the founding members set this forum up, we decided to use a commercial Forum package (there was a one-time purchase, plus an annual maintenance fee), because this customer funded business model offers us a better feature set and continuous through life improvement.  We chose the forum app from the company now called Invision Community.  This a PHP application that runs on a LAMP stack on a hosted Virtual Private Server (VPS).  The VPS has grown in cores / RAM / storage over the years as the size of the forum and its user population have also grown.

 

Three years ago, I moved the LAMP stack into a Docker Compose project, and the VPS runs this single dedicated Docker Project which is on open Github.  If you want to know more then read the README at Github: TerryE/docker-buildhub which explains the setup.

 

Our Hosting Provider is currently doing an infrastructure refresh so we have to migrate to a new VPS and at the same time I am retiring the use of Docker and switching to the open-source equivalent, podman.  See this issue, if you want to know more: A Gemini Conversation about Migrating VPs and switching to podman.  Warning: it's a bit long (17 pages) and technical, but it lays out the issues and context.

 

This topic is just to keep interested members informed.  Any technical discussion / comment is welcomed.

 

I have opened a companion topic in Forum News and Site Issues: Rehosting the Forum Virtual Server to discuss the more general / policy aspect to allow this thread to focus on the nerdy stuff. 🙂

 

Can I ask you all to keep to the technical discussions here and use the companion topic for general / policy stuff.  @TerryE is my user account but I also have the Godlike @Admin account  to hide / delete off-topic posts if the mods don't move them to the general topic.  🤣😱

 

Edited March 2 by TerryE
Addendum about technical only here

[MikeSharp01]

On 02/03/2025 at 13:52, TerryE said:

Our Hosting Provider is currently doing an infrastructure refresh so we have to migrate to a new VPS and at the same time I am retiring the use of Docker and switching to the open-source equivalent, Podman.  See this issue, if you want to know more: A Gemini Conversation about Migrating VPs and switching to podman.  Warning: it's a bit long (17 pages) and technical, but it lays out the issues and context.

Like the conversation posting - especially as you work to cajole it to focus on the question at hand! Switching to Podman makes perfect sense, it's open source so will probably evolve faster than Docker - who I see are fighting a solid rear guard action in the new world of contested containerisation, and it uses the docker command line which aids learning. How much, given what you have Git, of it will port straight forwardly and how much re-working is needed, outside anything associated with moving to a differently resources VPS.

[TerryE]

@MikeSharp01, podman in essence has two modes full native podman and docker compatibility mode where docker compose projects are configured the same but the execution is in a user mode account.  There are niggle and bumps as you can hide the fact that you are running rootless entirely in the host environment, but I have the stack up and running, first cut.  The performance hit for using a container approach is a couple of % at most and having the entire stack under git just has SO many pluses.

 

As far as working with AIs. I actually prefer Deepseek, because its chain of reasoning approach gives a more systematic approach and analysis, but its a bit overloaded ATM.  However, Genimi is a standard on a Chromebook and it's fast.  The issue is that by default its a bit patronising and treats the user as an idiot, but to be fair to Google probably 95% of its interactions are with idiots.  It's definitely a knack getting the AI to kick into expert mode.  The risk is that these AI are sincere and convincing but maybe 10% of the time the advice is flawed, so you always need to check -- trust but verify.

[Sparrowhawk]

I've just switched to Podman Quadlets with one client (single server setup) and liking it - yes it's not as nice as Docker Compose for me to write, but just having everything systemd is nice. Plus, rootless containers too.

 

Have we a CDN in front of the server for non-logged-in traffic? Again, something I do at $DayJob to reduce the size of server required.

[TerryE]

@Sparrowhawk, no CDN at the moment though the ICforum app does support using CDN.  This would drop our Apache requests and network traffic by maybe 4× though not our processing load as this is 90% PHP-FPM and D/B load.

 

As we talked about offline, I agree that it makes sense to stay with docker / compose for now.  I currently use Alpine for my docker images but having looked at the CLIB vs MUSL performance issues, it makes sense to switch to bullseye-slim.  I will ping you on the Github issue when I raise it. 

[TerryE]

I have just posted a new scoping issue on GitHub: Rework to Support the New VPS Environment .  My thanks to @Sparrowhawk for his input and advice.  After some discussion, we've decided to stick with Docker for now.  There is just too much else to be done to try to fit this in the Migration time line.

 

 

 

Edited March 7 by TerryE

[TerryE]

I've just updated companion Rehosting the Forum Virtual Server topic.

 

TBH, It didn't help that Heart Internet offered us a reduced spec VPS (less RAM, less SSD, less performance) for the same price, and this has involved quite a bit of pissing around / effort on my part to mitigate this.  But as @Sparrowhawk commented: the price that we're paying is really at the bottom range of what you'd expect to pay for a 6-core VPS of this spec.  You get what you pay for, I guess.  We try to keep the annual costs as low as possible to keep the forum free for all, and unencumbered by the need to advertise.

 

We'll just have to keep a close eye on actual performance post cut-over to see if we have any noticeable performance issues.  The GitHub issues give all of the technical details, if you are interested in more details. 

 

Thanks to @Sparrowhawk, @MikeSharp01, and @AliG for your support.

[Mr Punter]

Thanks very much for all the work you have put in to keep BuildHub running smoothly over the years.  Ad and spam free is just so refreshing.  Well done @TerryE!

[SteamyTea]

So have others had trouble logging on today?

[TerryE]

46 minutes ago, SteamyTea said:

So have others had trouble logging on today?

 

Yup, the forum went offline. Some gremlin still TBD.  We are peaking at something like 5 topic views per second during peak hours these days.

[Pocster]

14 minutes ago, TerryE said:

 

Yup, the forum went offline. Some gremlin still TBD.  We are peaking at something like 5 topic views per second during peak hours these days.

Yeah couldn’t login . Pornhub was ok though .

[SteamyTea]

8 minutes ago, Pocster said:

Pornhub was ok though

Plenty of jerky videos today was there?

[TerryE]

8 minutes ago, Pocster said:

Pornhub was ok though

I am surprised that you aren't into sextb.net or isn't your Japanese good enough?  😂

[Pocster]

2 hours ago, TerryE said:

I am surprised that you aren't into sextb.net or isn't your Japanese good enough?  😂

Not heard of that - but guess what I’m going to look at now !

[LiamJones]

Hey @TerryE. Google search results for buildhub are significantly degraded, I think since the above migration. Doesn’t look like Google is able to crawl the forum as well. https://support.google.com/webmasters/answer/7489871?hl=en#zippy=

[TerryE]

@LiamJones Thanks.  We'll look into this.  I've noticed that our guest views are down. I did tighten the robot.txt rules quite a lot to minimise the load from the AI bots, but from what you suggest we've gone overboard.

[TerryE]

I've found it. The sitemap.php module is adding an extra /ipb in the forum uri's which gets snotted up in our URIs, e.g.

• https://forum.buildhub.org.uk//ipb/sitemap.php?file=sitemap_content_forums_Topic_21

instead of:

• https://forum.buildhub.org.uk/sitemap.php?file=sitemap_content_forums_Topic_21

 

Rather than change the sitemap module, it's probably easier just to add a rewrite rule to remove the extraneous /ipb.  That way I don't need to mode the ICsuite code

 

There's an ACP option one the SEO tab to fix this. 🙂  @Liam Jones thanks again for the heads up.

Edited Wednesday at 12:22 by TerryE

[TerryE]

I have made the ACP change and sitemap.php seems to be giving the correct URLs now.  So let's see if Google picks this up.  The change that I added to robots.txt was to add the sitemap directive. Before that google was just indexing from the homepage.

[TerryE]

@Sparrowhawk, methinks this is your forte rather than mine.  Care to take a look and analyse the Google Search Console.  Thanks 

 

It might be still worth adding rule to remove the gash /ipb from the URL.  IIRC in the very early days the DocRoot didn't include the ipb directory so the rewrites added it, but on some VPS move we just made the DocumentRoot /var/www/ipb (see service/apache2/conf/extra/ssl.conf).  This predates the move to Docker since the change isn't in the git logs.

Edited Wednesday at 12:47 by TerryE

[Sparrowhawk]

@TerryE Yes let me know the credentials for our GSE account and I'll see how our account looks.

[TerryE]

10 hours ago, Sparrowhawk said:

@TerryE Yes let me know the credentials for our GSE account and I'll see how our account looks.

 

I think that I've worked out what is going on.  The GSE property was set up years ago when HTTP was our default and /ipb/ was in the URL. I've just added https://forum.buildhub.org.uk/  to my GSE account. If you WhatApp me your Google creds, then I'll add you in Users and Permissions for this site.

[MikeSharp01]

Should we add this wrinkle, not the details but the fact of where the GSE account is, to the Git site?

[TerryE]

@Sparrowhawk, I'll add you as soon as I've those creds, ditto you @MikeSharp01

 

11 hours ago, MikeSharp01 said:

Should we add this wrinkle, not the details but the fact of where the GSE account is, to the Git site?

 

We do have an issue about where we store all of our secret stuff like creds than is excluded from our public repos.  Maybe it's worth rolling it all up into an encrypted 7z file and putting this on our cloud server. That way we only need to share one password.