Duy Do

A Production Celery Worker Freeze Caused by epoll

Wed, 28 Jan 2026 10:03:00 +0700

Back in mid-2019, while I was working on the Data Platform team at Krom, I ran into a problem that lingered long after it was fixed. Our Celery workers would occasionally stop processing tasks without any obvious signal.

Supervisor still showed them as RUNNING, with uptimes measured in days, but zero tasks were processed. Data pipelines quietly stalled, alerts never fired, and the only clue was an occasional Broken pipe buried in the logs.

Honored to Be an Elastic Contributor

Wed, 05 Apr 2023 21:08:04 +0700

I’m thrilled to share that I have been recognized as an Elastic Bronze Contributor for 2023. This achievement is a reflection of my commitment to the open-source community and a testament to the valuable work being done in the Elasticsearch ecosystem.

My efforts have primarily focused on solving real-world challenges faced by developers in the Vietnamese market and beyond. My key areas of contribution include:

Vietnamese Analysis Plugin: Ensuring robust and accurate search capabilities for complex Vietnamese language data.

Understanding Elasticsearch searching data flow

Wed, 11 Jan 2023 22:32:29 +0700

Search is a generic term for information retrieval. Elasticsearch provides various retrieval capabilities, including full-text searches, geo searches, range searches, scripted searches, and aggregations.

How does Elasticsearch execute a search query behind the scenes? The following diagram shows data flow for searching operations.

Understanding Elasticsearch deleting data flow

Tue, 10 Jan 2023 12:51:30 +0700

How is a document deleted from Elasticsearch? The following diagram shows data flow behinds the scene of deleting a single document.

Understanding Elasticsearch indexing data flow

Mon, 09 Jan 2023 07:52:35 +0700

When a document is stored in Elasticsearch, it is indexed and fully searchable within 1 second. Elasticsearch uses an inverted index data structure that supports full-text searches efficiently and very fast.

An algorithm for finding right-truncatable prime numbers

Thu, 01 Aug 2019 12:10:30 +0700

Inspired from my status on Facebook:

The prime number 73939133 is very special, if removing each digit from right to left of that number we get another prime numbers: 7393913, 739391, 73939, 7393, 739, 73 and 7.

One of my friends gave a challenge:

Can you write a program reading a number N from keyboard then finding the nearest prime number to N that satisfies the characteristics of 73939133?

As promised him, I would solve this challenge.

Java: Type Promotion

Mon, 06 Aug 2018 18:51:46 +0700

Cách đây cũng khá lâu mình đem đoạn mã đơn giản bên dưới hỏi một số bạn làm Java, cả juniors và seniors, các bạn đều bảo rằng đoạn mã này không có vấn đề, biên dịch và chạy được.

byte b = 1;
b = b * 2;

Thật ra thì đoạn mã trên không thể biên dịch được.

Trước tiên thử phân tích xem vì sao các bạn trả lời như vậy. Nhìn vào đoạn mã thì thấy b được khai báo kiểu byte và gán giá trị 1, khai báo này hoàn toàn hợp lệ vì tầm trị của kiểu byte trong Java từ -128 đến 127.

Kỹ năng mềm

Tue, 05 Sep 2017 09:41:07 +0700

Mình không biết các ngành nghề khác thế nào, chứ riêng với ngành kỹ thuật phần mềm của chúng ta, kỹ năng mềm (soft skills) là yếu tố cực kỳ quan trọng, ảnh hưởng trực tiếp đến sự thăng tiến và thành công trong sự nghiệp.

Ngoài kiến thức chuyên môn, các bạn sinh viên hay các bạn mới vào nghề hãy nghiêm túc trau dồi kỹ năng mềm cho thật tốt, đặc biệt là kỹ năng giao tiếp. Thực tế cho thấy, khả năng giao tiếp, trình bày, và hợp tác ảnh hưởng trực tiếp đến hiệu suất công việc, khả năng lãnh đạo và tầm ảnh hưởng của bạn trong tổ chức.

Interview at Elastic

Thu, 20 Jul 2017 09:48:56 +0700

Some months ago, I had an interview with Elastic for Java Software/Support Engineer position. The interview was an interesting experience I’ve ever had, Elastic is a brilliant company with great people and amazing culture.

Alan Turing

Mon, 17 Jul 2017 19:01:54 +0700

Ngồi đợi con học, ghi chép đôi dòng về Alan Turing.

Alan Turing

EVM - Elasticsearch Version Manager

Sat, 20 May 2017 17:50:35 +0700

As an Elasticsearch consultant, I often work with different versions of Elasticsearch. To make my developer life easier, I created evm. It allows me to install, remove, test multiple versions easily.

How to build Elasticsearch Vietnamese Analysis Plugin

Fri, 21 Apr 2017 00:00:00 +0000

Recently, I’ve received many requests to build the Vietnamese Analysis plugin when a new version of Elasticsearch is released but sometimes I’m not available to do it immediately. In case of urgent, you can build the plugin yourself with following steps.

How Elasticsearch cluster works

Mon, 24 Oct 2016 11:16:00 +0700

This post is part of a series covering the architecture of Elasticsearch based on my experience while working with it. In this post, we’ll be discussing how the cluster works, try to find answers for following questions:

How a node in cluster talks to others?
What happens when a node joins or leaves the cluster?
What happens when a node stops or has encountered a problem?

Elasticsearch Consulting

Tue, 18 Oct 2016 16:54:21 +0700

With a decade of professional experience working with Elasticsearch and deep understanding of how it works internally, I’ve helped many customers successfully with Elasticsearch-based solutions. I offer followings consulting services:

Architect search applications based on your specific use case, data volume, growth needs and business requirements.
Optimize queries, index mapping and tune relevancy for existing Elasticsearch implementations.
Tune, scale Elasticsearch clusters to meet your needs.
Migrate legacy search engines to Elasticsearch.
Develop Elasticsearch plugins.

What Customers Are Saying

Dang Phan, Product Manager at GO1.com

Effective Persistence of Enums in Java: A Deep Dive into Safe and Scalable Design

Sun, 24 Jan 2010 08:17:50 +0700

How to persist an enumeration effectively? Here is my experience.

I’ll use Wordpress as an example. We all know that every post in Wordpress blog system has a status with possible values: draft, pending-review, published and so on.