Hello once again! It seems I've settled on a quarterly (or worse) cadence for these, but to be fair, I have been very, very busy. Let's go over all the new things going on, then we'll get to the fun numbers.
Updates:
- Cryptpad.lain.la was launched as a fully functional Cryptpad instance. This one is completely free of charge, as usual, and has just as much storage per user as the PAID tier on the "flagship" Cryptpad.fr instance.
- I now have a Minio S3 compatible object storage instance available and ready. This one took a lot of time to get right, but it's working like a charm now, using the same cached, quad-redundant architecture as Pomf. This service is by request only, but I haven't denied anyone's request. Email me at 7666@lain.la for an account!
- I now accept donations via multiple new, non-crypto methods. Please see this page if you're interested. I also wrote an article describing how that went, and how to prevent leaking your personal information when doing outbound or inbound donations.
- I trashed that LVM cache system I built at the start of the year for the Pomf cache. It just didn't work out. You can see why here. Performance is much improved now, although my WAN bandwidth is up a bit as well.
- I have outlined my free hosting offer in complete detail here, so that I don't get asked the same question twenty times.
- In the hardware department, I have a brand new hot spare storage server ready to go, just in case the primary one craps out. It's as simple as just yanking all the drives out of the old server and stuffing them in the new one. Image below! It's another 3.5" chassis R730xd with the internal 3.5" expansion bay addon.
- I added Keybase to my public presence. You can see it here. I don't intend to use it very often.
- I built a brand new PKI for my OpenVPN deployment. It includes an offline CA (no internet access at all) and individualized certs per tunnel, rather than using the same cert for them all. This is a far more secure PKI to work with going forward.
- There has been some instability with my OpenVPN tunnels. I'm not sure why yet, but this new auto restart system, modeled off the failover system, works like a charm and will get us through until I find a permanent fix. You can read the code here, each node runs its own version pinging its proper gateway.
Metrics:
- Cluster stats:
- VM Count: 49 on-prem, 17 hentai nodes, 7 endpoints, for a total of 73 VMs. Dang! This is a 50% increase over March's metrics!
- Edge Bandwidth: This has chilled out a little bit, but we're still pushing an impressive 605TB a month! Here's a graph of one of the four main edge nodes. Multiply by four to reach the real total, then add a little bit extra for the rest of the infrastructure.
- Internal bandwidth (WAN) is where we take a large hit in. With the caching architecture rebuilt, and with a lot of new uncacheable traffic, these numbers have tripled. Again, I now have a business line, but this is still a pretty crazy amount of ISP traffic.
- Costs: These have gone up a little bit since the last article, mainly because I now factor in environmental costs such as cooling. I also have more NVMe storage to pay for now. I reduced the costs a hair by factoring in some small donations as well as some cheeky credit card point harvesting tricks. The total is $561.00/mo.
- Uptime: Still pretty good! Most services are almost at 4 9's of uptime. See here for details.
So, that's all for now! As always, feel free to email me at 7666@lain.la to say hello or if you have questions or suggestions.