Shubham Shingare
374 posts

Shubham Shingare
@_shubham_s19
I want to learn cyber security , Ethical hacking😎 interested in Artificial intelligence 👑
Katılım Ocak 2022
240 Takip Edilen25 Takipçiler

@rbkayz Getting the desired output is not enough, code needs to be light and flexible.
English

We had a fire on prod today! 🔥
A lot of people who aren't deep into the tech don't really get to see what really happens.
So here's the story
1/ At 5:30 AM UTC today, our database CPU usage started to spike
Normally, some spikes are expected and it goes back to normal fast.
But when it consistently remained over 80% for more than 2 min, alarms hit our WA group and slack
2/ My engg team and I got on a call to begin the triage. The actual process to treat is very much like a medical ER. First step is checking the vitals
- Check critical front-end operations
- Check all our queues for any buildup
- Check for spikes in latency or error logs
Generally, you'll get some sense of what is the likely issue
Today was a bit anomalous. Errors were low, BUT everything was just slow.
Queries were taking 4x the time on average, which was causing a case of problems - build-up in queues, sluggish FE, etc. And we work on WA messaging, so even small latencies feel annoying
3/ Since we couldn't determine the cause, we needed to treat the symptom, which meant alleviating the pressure on the db
At this point, we needed to start making some serious calls i.e. shutting down non-critical services. It'll cause some pain for folks, but manageable (e.g. you shut down triggers, analytics, or non-core AI ops like summarize or translate)
There's a defined cascade you follow to reduce the number of db queries
It helped a bit, but didn't really make a difference. CPU stayed well above 80pc
Now, I start to get worried because we don't have a line of sight into a timeline for the solve.
We raised the severity of the incident, and call an all-hands (sales as well) to manage customers
4/ Then we started getting deeper into the db's active statements. And then we figured out the cause
Postgres has an autovacuum function that cleans up dead tuples (discarded rows during updates or deletes)
And one specific table (which is many many GBs) had a high influx of dead tuples due to a large spike in signups we've been having at @periskope_app
So @supabase's autovacuum had kicked in to clean up these tuples, but the infra was not sufficient for it to actually scan the table and complete
Each time it would attempt to scan the table, it would spike the CPU and it would fail. And this cycle repeated endlessly
Sadly, killing that process was not an option because this was a type of autovacuum called a "wraparound", and killing it can causes critical failures on the db
So the only option left now is to scale up the compute
5/ Scaling up the compute is NOT something I want to do in the middle of peak hours. It's a full downtime and not a good look
But these were desperate times with no other option. So we initiated an emergengy maintenance window
The good part is that we've prepped for this, so there's a clear process
First you raise the incident level on our status page. This automatically takes down our front-end to a maintenance state.
Then you scale redis as a temp backup store for incoming events, and shut down the db for the compute increase
6/ The upgrade takes about 2 min. But that's a very very long 2 min where all you can do is wait, and hope it goes fine
Thankfully it did, and then CPU started to normalize. Queries returned to normal latencies, and we can restore the FE
This kicked off a post-op process, which involves data integrity, writing a RCA, and most importantly, fielding customer questions
Days like this happen more than you know. This was a highly complex and technically challenging day at work.
But there was actually very little code that was written. Just a lot of tweaking and tuning. And some critical decisions that require experience to take (no book is going to teach you this)

English

@Aishi_ki_taishi @callmeveizir Samajao yarr tumhare ladle ko
Filipino

@callmeveizir @_shubham_s19 What is this behaviour munnu. Reply karo grp pr
English

@akshaymarch7 Bro just created a girlfriend to explain how backend things work.
English

Yesterday my girlfriend asked me a simple question:
“How does food ordering apps like Zomato know exactly when my order is going to arrive?”
I laughed and said, “Magic.”
But she didn’t buy it.
So I explained…
Every time you place an order:
0. Your location gets sent to the backend.
1. The restaurant confirms prep time.
2. A delivery partner is matched in real time.
3. Traffic and distance are calculated.
4. And every 2 seconds, the app updates your screen with the rider’s live location.
Now imagine this happening for millions of orders at the same time across India.
That “magic” is NodeJS.
Because Node can handle thousands of concurrent connections without breaking a sweat, it’s the secret behind fast, real-time updates we take for granted.
The fun part? Once I explained this, she just said:
“Okay, but will the food still come in 20 minutes?” 😂
That’s when I realized, users don’t care about backend tech. But as developers, we must. Because NodeJS is literally powering the apps we live on daily.
And if you’re serious about backend, learning NodeJS deeply is non-negotiable.
That’s exactly why I created Namaste NodeJS, to take you from “just writing APIs” to understanding how real systems similar to like Zomato, Swiggy, and Netflix actually run behind the scenes.
👉 Check it out here: namastedev.com/learn/namaste-…
Cheers,
Akshay Saini 🚀

English

@callmeveizir @Aishi_ki_taishi Toh kya thread me hi discuss krle sab😅
Indonesia

@_shubham_s19 @Aishi_ki_taishi 🤣🤣🤣 bro full on twitter aur kon daddy bhai kuch naa🤣
हिन्दी

@callmeveizir @Aishi_ki_taishi Munnu ye kya dekhta hu me... Gpr me msg ka reply nhi diya jata terese .. aur yha logon ka daddy bna ghum rha hai ...?
Indonesia

@Sartaj4congress @mrsinha This post is basically making fun of people who are trying to destroy India , but still @Sartaj4congress call our own people andhbhakt.
On which basis you called andhbhakt?
English

CEO of Alphabet Google 🇮🇳
CEO of Microsoft 🇮🇳
CEO of YouTube 🇮🇳
CEO of Adobe 🇮🇳
CEO of World Bank Group 🇮🇳
CEO of IBM 🇮🇳
CEO of Albertsons 🇮🇳
CEO of Infosys 🇮🇳
CEO of NetApp 🇮🇳
CEO of Palo Alto Networks 🇮🇳
CEO of Arista Networks 🇮🇳
CEO of Novartis 🇮🇳
CEO of Micron Technology 🇮🇳
CEO of Honeywell 🇮🇳
CEO of Flex 🇮🇳
CEO of Wayfair 🇮🇳
CEO of Chanel 🇮🇳
CEO of Cognizant 🇮🇳
Note: citizens of India, citizens of other countries but born in India, Indian origin, or with Indian parents
English

@aaraynsh @NalinisKitchen Bro even sam Altman fought with elon on twitter.
English

@NalinisKitchen First step of going desi is deactivating Twitter account 🥲
English

@samitkapoorr @aceternitylabs He needs the HR department too.
English


@bh_saksham27 @AravSrinivas @perplexity_ai @airtelindia Jio people will get a 12% hike .
.
.
.
In recharge price
English

@callmeveizir Hamare time pe aisa hi hota tha...
Laptop leke aptitude kon solve krta hai be
Eesti

Hi folks!
Need some solid financial advice
If someone earns around ₹8–10 Lacs per month, how should they manage their money?
What’s a good way to split between spending, saving, and investing?
Would love to hear how you’d approach this.
#bugbountylife #bugbounty #finance
GIF
English













