I don't know whether i will be able to solve it. But I love to read to about problems and solve them. Maybe someone can help you with solution or a hint which you are looking for. But i would like to atleast think about the solution
Here's one - Timeout Issue
Architecture: Public ALB -> golang server on kubernetes
We have a Golang http server. It collects form-data from the POST requests it gets.
The form-data can be at max 500KB
In the golang code, we're using net/http for the server. And while reading the data from the request we use the FormFile().
Some of our requests time out (read timeout) at this juncture - while reading the data.
For testing I'm using the same file data in the POST request to the server. For some requests, the FormFile function takes <1s, for others it hits the read timeout (set to 5s). This happens even without much load, CPU and Memory look good for the pod and the k8 node.
I can increase the read time timeout say to 10s and it saves some of the requests orm timing out, but then we still have a few timeouts here too.. so we have to with the root cause before we increase the timeout limit.
Why the same file takes 1s some times to be read by golang were as 10s other times, even when there is no resource pressure?
Golang server is running as a pod. I've tried increasing the number of pods, cpu limits of the pods. Not useful.
I would see what sort of http protocol we are using to make that request- http1.1,1.2, or 2.x. I faced similar issues with AWS ALB earlier. Try changing http versions from 1.x to 2.x
Airflow DAG slows to a crawl on month‑end - We partition raw events into daily S3 folders. The PySpark job that compacts a full month runs fine days 1‑29 (~15 min) but explodes to 2 hrs on the 30/31st. Tried salting, tried dynamic allocation.
My hunch: small‑file syndrome + skew. What’s your go‑to fix without doubling cluster size?
A few questions -> Are there certain activities which are exclusively being done on 30/31st ? (For example accounting and back calculations)
Is the hdfs file size roughly same on a day to day basis ?
What’s your methodology for pushing raw events into a file ? (By batch size+ wait period, or something else ) (this usually leads to a lot of files getting created if done improperly)
Why not do daily snapshots followed by weekly/monthly snapshots ?
Now we are talking the real problems rather than a 20 y/o crying for girls. Thanks everyone . We are fucking brilliant minds solving crazy problems. I fully respect everyone who have problems and share here. But I hate to the core who just cribs even if you tell them the solution
Problem: There around 150k ish people who will be constantly changing their position. Ex: Promotion cycle in a firm, I need to come up with a solution wherein all these users get their mailbox either upgraded or downgraded and their licenses getting correctly assigned to them and when they do get those downgrades I have to migrate their individual accounts to a litigation hold in case they lose any important emails while keeping the performance of the mail servers not going down.
If anyone can share and discuss about something interesting you might be working on, tricky to solve. Feeling a little bored with all the routine stuff.
Will entertain only serious questions that respect anonymity. I don’t mind humor, but trolling is strict no no.
Conducted one AMA session last week & was overwhelmed by both number & positive feedback. I might choose to skip any questio...
Basically let's say if you're on remote and you are working on a large code base which was built over the years. You have been assigned something to do & you end up in a hurdle or loop. You can't seem to get it working or whatever you th...
I am a recent graduate with a BTECH CSE degree working at a very early age start-up as software developer this seems very normal but now that's a catch I am working from past 5 months and now I am started feeling stuck and overwhelmed li...
