2024
Fleet
The 2am grindset is real! We got the ghost stories playing in the background, which describes exactly how I feel when operating in kubernetes! After we execute the manual SQL queries, we need to wrap back around and update the shared extensions. There is another point of error that I have to look into, which is the pgnet because I believe the cURL that it requires is only in ubuntu 24 and not ubuntu 20. Hopefully the 15.6.x postgres image that Supabase plans to release or 16.x postgres image hopefully is built using nix and ubuntu 24.
Shared
2:28am - Preload Error
Looks like that the shared_preload_libraries did not infact load the pg_net, so we will have to go back around and fix that.
Furthermore, we can also use this part to load the kilobase extension too, okay made some changes to the chart to include the extensions.
Kilobase Extension
Our custom extension did not seem to load and thus we need to go back and change the build. I believe this should be an easy issue to resolve because we need to remove the OpenSSL part of the reqwest client. Okay there was a small step that we skipped, which would be to bump the version control up, making sure that we are using the latest version.
When doing the version bump, we need to edit a couple files and in the future we should have a shell script that would make these changes for us during a pull request. Maybe after an official image is published, have it go through and make changes to all the files.
- ”Cargo.toml” under /apps/kilobase/
- “project.json” under /apps/kilobase/
- “values.yaml” under the /migrations/kube/charts/kilobase/supabase/
Keeping the kilobase version consistent across all three files will be important.
Okay the shell script version_bump.sh handles this.
Next error is in the misisng libc6, which I am assuming the linking was broken? Let us go ahead and add it into the Dockerfile and see if it helps solve that problem. Damn it, the error was not because of a linking problem but rather a mis-match in the operating system for building the extension on ubuntu-24 and running that so file in ubuntu-20. I been moving around different ubuntu versions and forgot that ubuntu-24 is definitely the wave. Ugh, here are the rest of the notes related to this.
Okay there are two ways we can fix this issue, we can fix the action that runs the extension builder using ubuntu-20.04 instead of ubuntu-latest?
Or we can move the extension building back into the Dockerfile as we had it from the start, ugh, that means we need to make a couple adjustments to our overall build process.
The first step in setting up the docker build would be to directly reference the jedi package via the crates rather than a local path.
That would split the updating for the jedi package up a bit but for now it should be okay, I can sense the future problems that this might cause with our locks but all of this is part of the learning experience.
We will shift the Cargo.toml from :
jedi = { path = "../../packages/jedi" }
# TO
jedi = "0.1.14"
Actually decoupling the direct reference to jedi and maintaining the jedi workflow will be a better way for our future development cycles.
Now this also means that we need to adjust the Dockerfile next, preparing the STAGE 1 again for building out the kilobase extension.
Docker
Dear lord building docker images on Windows WSL is terrible, I am honestly waiting forever for docker builds and some of the errors are just being skipped. I guess I been spoiled with orbstack and some of the newer tech, that going back to Docker desktop feels like entering dev cycle hell all over again. Going to version bump the whole container and see what the errors are within the dev branch, then go back around and compare the errors from my end? That would mean we have to comment out some of the dev ci workflow as well because we moved all of pgrx into the Dockerfile, but it might make sense to test it before going into the Docker buildx. Both of these options would make sense in this situation, testing the extension building before doing the full docker build, maybe within the dev cycle we will keep the extension tests. Then inside of the alpha branch, we will do a pgrx build and then a docker build test, this way we can keep both of the tests but split them up a bit between the two branches.
GLIBC
- 7:00pm
Error log from the kubectl logs supabase-release-supabase-db-1-initdb-flnzv -n supabase
{
"level":"info",
"ts":"2024-09-25T22:16:10Z",
"logger":"pg_ctl",
"msg":"waiting for server to start....2024-09-25 22:16:10.661 UTC [29] FATAL: could not load library \"/usr/lib/postgresql/15/lib/kilobase.so\": /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.33' not found (required by /usr/lib/postgresql/15/lib/kilobase.so)",
"pipe":"stdout",
"logging_pod":"supabase-release-supabase-db-1-initdb"
}
Keeping this error in the notes as a reference as I know we will have this issue in the future again when we have to rebuild the supabase image using the ubuntu 24 / postgres 16.
NVDA
Ending up selling all the Nvidia call options and preparing to move some of the cash into CDs before the rates start to drop! This will be a good way to save some extra cash that I will have to pay the IRS anyhow, might as well have it sit and collect some interest before the start of 2025. So profits, about 50%, will be moved into a CD that should end right before April 2025, then we pay the IRS with that large sum, overpaying them and having them deposit back the excess cash into a T bill account. One of my goals would be to get around six figures in T Bill funds before the age of 40, this would be then used as an extra extra extra emergency fund for the future, granted the rate of return on the t bills are pretty trash but all of this is just part of that experience.
2023
- 10:05am - I completely forgot about the work backlog, so I need to focus on catching up on that end. I will dedicate a couple hours to just getting myself ahead of the backlog? I could mix it in with the other backlogs, so that I do not get too bored of handling issue tickets.
- 11:17am - The storage of the meta engine should really be done on Amazon or some sort of external cloud provider, I found that the metadata being on AWS was the best option in this hybrid cluster setup. The future really seems to be a combination of a hybrid cloud, where a decent amount of the important meta-data would be stored in the cloud while the heavy functions and files are held within your own instance.
- 4:22pm - Scaling the YoRHa UI down to just around 400px~ and keeping it there for now. This way its mobile focused at first before I migrate it be desktop / landscape friendly. I might shift out of the UX/UI area of the widget and focus more on the backend and Appwrite integration for the time being. I suppose these notes are just for getting a better understanding from my end.
- 6:36pm - Okay I was able to swap out the general characters for the Game! I am thinking of keeping it simple for now and only three playable characters before migrating over to a bit more advance style features. There are a couple other aspects that I wanted to overview.
Quote
Love is the only force capable of transforming an enemy into a friend. — Martin Luther King Jr.
Tasks
- [ ]