SuperQue

9 ms·

31.
by SuperQue 4y ago
2. Treat reliability as if it is its own feature. It's also "Treat reliability like a software engineering problem not a process/operations problem".
32.
by SuperQue 4y ago
No, it really does have to do with organization priorities. You can make things reliable at scale with proper processes and automation.
33.
by SuperQue 4y ago
You seem to have missed the point of what makes Kubernetes a really nice system to work with. It's not the machines. It's the comosible components, the service building blocks I can get with an API call. * Want traffic into a clus
34.
by SuperQue 4y ago
Not to mention all the various in-house systems in the 2010-era. I worked at a small place back in 2013 that had built their own container platform about a year before Docker was released. It was more like workqueue than borg. But it worked
35.
by SuperQue 4y ago
Pretty much this. Although it does take some good physical senses to really pull out the spices/herbs in some things.
36.
by SuperQue 4y ago
I work for a large online service. We barely need your IP to track you. There are _so_ many other variables sites can use to track you. Even when you switch networks completely. You're fear of IP tracking is outdated.
37.
by SuperQue 4y ago
I hadn't been to Chicago in a while, but I ended up taking a taxi from the airport due to how slow the trains were I've been living in Germany for quite a while. It's about the same distance from my home to the airport as O&#
38.
by SuperQue 4y ago
We added it to the Prometheus node_exporter ages ago. It's extremely useful in detecting system stalls. What I really need now is to have this information at the container level in cgroups v2. https://github.com/google&
39.
by SuperQue 4y ago
Go is the prime example. For example, in your typical web application a Goroutine is spawned for every incoming connection. This basically gives you a dedicated runtime for each simultaneous connection. In Node or Python, this would be the
40.
by SuperQue 4y ago
The problem is there is more than one kind of "fast". There is latency "fast" and throughput "fast". For a lot of web stuff you need throughput more than you need latency "fast". In order to do throug
41.
by SuperQue 4y ago
Berlin to Hamburg (~1:50) is a nice route. There's almost nothing in terms of population between these two cities. However, every 2 hours there is a train that stops at Wittenberge (population 8k) and Ludwiglust (population 12k). I ass
42.
by SuperQue 4y ago
We had 2 initial prototypes to index the bay area. I think there were maybe 10s of cars in the first big sweep over the US. I have no idea how many cars they're running these days.
43.
by SuperQue 4y ago
Street views was actually started as a Stanford project called CityBlock. LOLOL, business analysis? The idea was entirely an engineering nerd project. We wanted to index the physical world, just like the book scanning project and a bunch of
44.
by SuperQue 4y ago
Prometheus is in the process of developing a similar automatic log-linear histogram bucket type. The goal is to make it as cheap as a 10-15 bucket histogram, but not require pre-defined buckets. https://github.com/prometheus
45.
by SuperQue 4y ago
Windows users just look at CPU utilization, which is what most Linux users should be doing as well. Monitoring both total system and single core utilization. Of course, 100% CPU use also isn't a "problem", it's just a fa
46.
by SuperQue 5y ago
One interesting question I have is regards to global availability. With our current Thanos deployment, we can tie a single geo regional deployment together with a tiered query engine. Basically like this: "Global Query Layer" -&gt
47.
by SuperQue 5y ago
Yup, it's tiny and simple. But the goals are wildly different. The same Prometheus binary you can run on a Pi scales to millions of series and millions of samples per second. The Prometheus TSDB itself is only one part of a the larger
48.
by SuperQue 5y ago
Prometheus works fine as-is on Pi devices. You'll spend most of your memory on ingestion buffering. I did the same tests as you did a while back, it only saves like 25-50MiB of memory IIRC. The only thing you really need to worry about
49.
by SuperQue 5y ago
If only there was a monitoring system you could install to measure such a thing.
50.
by SuperQue 5y ago
It's a little bit of both. Simple to implement, simple (and fast/cheap) to parse, and human readable.
51.
by SuperQue 5y ago
Yea, there still some corner cases and implementation bugs out there. We spent months deliberating how to deal with some of these. Because the base libraries in some languages just don't produce string output from IEEE 754 the same way
52.
by SuperQue 5y ago
Sorta, I would say it's an evolution of varz. IIRC (it's been almost a decade since I used varz), having multiple label values would be a map of maps in varz. It got quite ugly if you wanted to have a number of dimensions.
53.
by SuperQue 5y ago
Prometheus supported a JSON representation in the beginning. It was deprecated and removed before 1.0. The current exposition format was created because it cut CPU and memory for scraping metrics in half. JSON, especially free-form JSON, is
54.
by SuperQue 5y ago
We've already talked about it at length, for many years. There's an approved development proposal to implement it. But, nobody has stepped up to contribute the code. The main problem is, Go doesn't have any kind of reasonable
55.
by SuperQue 5y ago
That's not running binaries works. Only the parts of the code that are in use are paged into the page cache. So if you only use a couple of the features, it fits in cache just fine.
56.
by SuperQue 5y ago
I've done this test in the past. The typical footprint of default Prometheus is around 100MiB of RSS. Removing everything but flie and static configs reduces it to about 50MiB. Interesting for more embedded use cases, but not really a
57.
by SuperQue 5y ago
You can use the Prometheus TSDB package, it's easy to import and use. Plus it's designed around mmap, so the kernel deals with the caching. You don't have to use a whole Prometheus server to make something like this work.
58.
by SuperQue 5y ago
It's not just that. It's that hate and lies have the same weight as any other opinion. Paradox of tolerance.
59.
by SuperQue 5y ago
I'm no expert in these things, but here's my understanding. AFAIK, it's up to copyright assignment for re-licensing. For example, contributing to Prometheus assigns all copyright to "The Prometheus Authors". The DCO
60.
by SuperQue 5y ago
The problem is CLAs. They are a much higher barrier to entry for a lot of contributors. No, they're not all the same, but they make it harder. Each CLA needs to be reviewed and approved by company lawyers. It's kinda like creating

More ›