Windows, why are you so weird? What on earth were you doing in the background to consume 100 GiB of RAM before eventually releasing it? It sure wasn't in use by any process 😬
| GitHub | https://github.com/ipsi |
| Blog | https://www.andrew-thorburn.com/ |
Andrew Thorburn
@ipsi
- 22 Followers
- 28 Following
- 104 Posts
Wandered over from Twitter (@nzipsi / @athorburn4); Stay a while and listen.
Huh. Sleep Apnoea in the Apple Watch? That's directly relevant to me, might actually consider the new Apple Watch 😱
I occasionally get scam texts, and it can be interesting to play along and see what they know and don't know. In this case, we're talking in German despite the person on the other side ostensibly being from England - does the scammer not know that NZ is an English-speaking country? 😂
One of my OG #HomePods decided to blow a fuse or something tonight - sudden popping sound, repetitive playing of a very loud sound, extremely hot to the touch, lovely smell of burning plastic. I’ve put it outside (on concrete!) the hope that it’ll cool off and not burst into flames.
If that doesn’t work, I might to organise a group call between the fire department, the bomb squad, and an exorcist 🤔😬
A lot of these frustrations stem from trying to set up two DNS servers (one primary, one backup, because I'm prone to breaking things 😬), and it would have been very nice if I could have just told Docker Swarm “Hey, I want two DNS servers, please instantiate them on two Swarm nodes using *this* VLan, and give them an IP from this pool which is exactly the same size as the number of instances. Also, when I want to change the config file, please do a rolling update, so I don't lose DNS 😅 “
* Changing the IP address of a machine breaks Swarm. Force-leaving the swarm and re-joining breaks networking for *every* container on the node, even if they're not using Swarm! They *all* have to be force-recreated, ideally in the right order.
* Non-swarm Containers can join Swarm networks… except for Docker Compose on a follower node (you can attach it after, but…😕 )
* Smaller bugs - even on leaders, compose can't attach to a Swarm network - another container has to do it first.
* Non-swarm Containers can join Swarm networks… except for Docker Compose on a follower node (you can attach it after, but…😕 )
* Smaller bugs - even on leaders, compose can't attach to a Swarm network - another container has to do it first.
I really want to like Docker Swarm - it's *very nearly* a simple alternative to K8s, but is sadly missing a few things I would really, really like:
* Can create config-maps! But they're immutable - if you want to change it you need to delete the associated stacks, delete the config, create the config, create the stack 😬
* Can't assign static IPs. I realise that's a bit awkward with a "create N replicas" thing, but something like "10.0.0.[1-4]" would be great.
If you really want to use VLANs, then at the very least stick to one per Ethernet port per computer.
Having multiple small, cheap computers is in many ways less painful than having one large, powerful PC and plonking a thousand docker containers on it.
Some of that is just Docker being an awkward fit for hosting lots of stateful apps. Part of it is that although the alternatives (e.g., Proxmox) are more flexible, they're *way* less plug-and-play, at least for what I'm doing.
If anyone has the desire to get into self-hosting, my advice (after a couple of fun evenings 😬 ) would be: keep your networking *simple* - flat network, no vlans. You will avoid *so many* hard-to-debug problems (like needing an mDNS repeater such as Avahi, so that my Home Assistant HomeKit bridge can be picked up by my iPhone/etc in the other VLAN).
I'd also like to give a shoutout to OpenSearch (and probably also ElasticSearch) for being probably the best log-search application while *also* the hardest to stand up (by a _huge_ margin), *and* the most brittle. It _really_ doesn't like running in Docker Swarm 🤔
Could be related to running the swarm nodes on different VLANs, so going to nuke it all and try again now that's fixed. If it still struggles to work, then I guess it's back to grep for searching logs...