After many years of just ‘taking’ I mature and upgraded my setup to be able to give community back what I took.
I setup RAID 6 FreeNas machine. Thanks to playing with those I understood why everybody in FreeNas community hate Realtek and why people just say to throw it away and get something that works (FreeNas: Realtek 8111F and re0: Watchdog Timeout Error and FreeNas: Realtek 8111F and “re0: Watchdog Timeout Error” – universal solution). In the end I join the ‘hate club’ and bought Intel PRO 1000 PT Dual Port NIC. Fun part is that Linux handle Realtek with ease but I been with Linux when It didn’t like most of hardware that Windows support so I can understand that there are still hardware issues with FreeBSD.
Cool, FreeNAS install without any issues.
If you unfamiliar with FreeNAS it comes with ‘plugins’ that are basically jails with pre-configured application that ‘just works’.
I slap that Install button and almost instantaneous I had working Transmission 2.93 (FreeNAS should work about pushing updates to plugins faster but yeah, nothing to complicate that we cannot fix ourself if needed)
I mounted needed directories, slap few torrent files that I already had there and enjoyed how transmission is handling all those sweet, sweet files out there.
And Tranmission was working and it was doing good. But after many, many torrent more that join the seeding happy farm not many people did connect to share the love from me. I started tweaking Transmission. Setting by setting I scope documentation looking for holy grail, some magic setting that would boost simultaneous connections. I found few. So I created new config file, and it was good. Not the best but I notice some connection increase. But I was thinking that maybe its because the application is inside jail aka virtual network interface. Few threads on popular forums, many many blog post later I saw the solution `allow.raw_sockets=true`. Great I saw a big improvement with connections – but who know if that was really all this tweaking doing or just someone join the swarm.
Few days later I notice big (I could even say tremendous) packet drop from/in my network. What was happening ? Router that is used between NAS and WAN have working QoS rules so it shouldn’t be a problem with using to much of bandwidth… So what was it ?
Culprit here was the CPU inside the router – we need to remember that using fancy feature that make router even smarter use CPU. QoS, Firewall, VPN all of them take some of that sweet-sweet pie (CPU). But why its dropping packets you ask? Simple, there is less and less CPU power to handle all the traffic that is going to and from network so everything is super slow. But wait, didn’t it work before Transmission? Yes it did – and we could disable some functions but we can do better. Lets optimize some firewall rules, we can cut QoS from 10 types to 5 without big sacrifices. Is that enough? No, but we can tweak both Transmission and firewall rule for it. Let set ‘Max connections’ to smaller number and double that with same (or 1% more on firewall rule so we are double sure that anything that is marked as ‘new connection’ and is about our limit get dropped).
And this gave enough room for that sweet-sweet pie named CPU could handle ‘few’ packets more.
And it was good for months and months…
Until the corruption start to show !