metadata

拿 fq 看一些二進位檔案的 metadata

前幾天的 Hacker News Daily 上看到「Fq: Jq for Binary Formats (github.com/wader)」這個工具 fq，可以對二進位檔案操作：

Tool, language and decoders for working with binary data.

Ubuntu 22.04 以後的版本有套件可以直接裝 (就 apt install fq)，不過 22.04 內的版本比較舊，可以考慮抓 binary 下來丟到 ~/bin 之類的地方用，不過目前應該是先丟著，等明年 24.04 的時候就會自己換新了？

官方範例的 fq . file 可以看 metadata，這樣還算是好用，不用自己 hexdump 對半天；另外「Supported formats」這邊有每種格式對應的 column name 可以讓你操作，參考 fq --help 的說明就可以拉一些欄位出來用...

AWS 的 VPC 內可以用 IPv6 的 Metadata Service、Time Service 與 DNS Server 了

在「IPv6 endpoints are now available for the Amazon EC2 Instance Metadata Service, Amazon Time Sync Service, and Amazon VPC DNS Server」這邊看到的，AWS 把一些基礎服務補上了 IPv6 interface：

The Amazon EC2 Instance Metadata Service, Amazon Time Sync Service, and Amazon VPC DNS server can now be accessed over IPv6 endpoints by instances built on the Nitro System.

Metadata Service 的部份在「Use IMDSv2」這邊可以看到，在 IPv4 的時候是 169.254.169.254，在 IPv6 時則是 fd00:ec2::254：

The examples in this section use the IPv4 address of the instance metadata service: 169.254.169.254. If you are retrieving instance metadata for EC2 instances over the IPv6 address, ensure that you enable and use the IPv6 address instead: fd00:ec2::254. The IPv6 address of the instance metadata service is compatible with IMDSv2 commands. The IPv6 address is only accessible on Instances built on the Nitro System.

Time Service 的部份可以在「Set the time for your Linux instance」這邊翻到，可以看到 fd00:ec2::123 這個位置：

The Amazon Time Sync Service is available through NTP at the 169.254.169.123 IPv4 address or the fd00:ec2::123 IPv6 address for any instance running in a VPC. The IPv6 address is only accessible on Instances built on the Nitro System.

DNS 的部份則是在「DNS support for your VPC」這邊翻，可以看到 fd00:ec2::253 這個 IPv6 address 的 DNS 服務：

If this attribute is true, queries to the Amazon provided DNS server at the 169.254.169.253 IPv4 address (or the reserved IP address at the base of the VPC IPv4 network range plus two) and the fd00:ec2::253 IPv6 address will succeed. The IPv6 address is only accessible on EC2 instances built on the Nitro System.

算是又補齊了一些基礎建設...

另外 EC2 剛好這三個字都在十六進位表示法的範圍內，就拿來當 IPv6 address 也是讓人好記很多。

另外結尾故意用 ::254、::123、::253 (要注意這是十六進位) 也讓人好記很多，而不是 ::fe 這類 hex。

Kafka 拔掉 ZooKeeper 的進展已經進到 trunk

差不多一年前提到了 Kafka 打算拔掉 ZooKeeper 的計畫 (在「Kafka 拔掉 ZooKeeper 的計畫」這邊有提到)，昨天 Confluent 發了一篇進展：「Apache Kafka Made Simple: A First Glimpse of a Kafka Without ZooKeeper」。

這件事情可以在「KIP-500: Replace ZooKeeper with a Self-Managed Metadata Quorum」這邊看到說明，另外在 Jira 上面則是在 KAFKA-9119 這邊追蹤。

目前的好消息就是 KIP-500 的程式碼都已經推進 trunk，目標在 2.8 釋出時就可以用：

So we’re very pleased to say that the early access of the KIP-500 code has been committed to trunk and is expected to be included in the upcoming 2.8 release.

文件裡面會用「Kafka Raft Metadata mode」或是「KRaft」稱呼這種模式：

For the first time, you can run Kafka without ZooKeeper. We call this the Kafka Raft Metadata mode, typically shortened to KRaft (pronounced like craft) mode.

然後有缺一些功能，另外就是常態性宣導這是測試版，建議不要在 production 環境用：

Beware, there are some features that are not available in this early-access release. We do not yet support the use of ACLs and other security features or transactions. Also, both partition reassignment and JBOD are unsupported in KRaft mode (these are anticipated to be available in an Apache Kafka release later in the year). Hence, consider the quorum controller experimental software—we don’t advise subjecting it to production workloads. If you do try out the software, however, you’ll find a host of new advantages: It’s simpler to deploy and operate, you can run Kafka in its entirety as a single process, and it can accommodate significantly more partitions per cluster (see measurements below).

目前丟出來的測試數據可以看到有大幅改善，但仔細看居然是停機與恢復時間：

不知道實際的效能上有多少影響 (正面或是負面)，還是得等了...

Android 版的 Facebook 蒐集大量系統資訊

Android 版的 Facebook 把整個 system library 的 metadata 都傳上去了：

Facebook scans system libraries from their Android app user’s phone in the background and uploads them to their server

This is called "Global Library Collector" at Facebook, known as "GLC" in app’s code

It periodically uploads metadata of system libraries to the server pic.twitter.com/olwk1BPMoQ

— Jane Manchun Wong (@wongmjane) August 30, 2019

而在 iOS 上，目前沒有觀察到類似的行為：

really interesting catch Jane. I have not seen this in the iOS version. my bet is that they want better stability on Android so want to analyze system library variants - and apparently did not even think to ask user permission.

— Will Strafach (@chronic) August 30, 2019

這些資訊足以識別裝置的單一性，而且難以更改，對於廣告有超大的幫助...

把 HDFS 的 metadata 丟上 NewSQL

把 HDFS 的效能瓶頸 metadata server 的資料改到 NewSQL 上使得效能大幅提昇：「HopsFS: Scaling Hierarchical File System Metadata Using NewSQL Databases」。

In this paper, we introduce HopsFS, a next generation distribution of the Hadoop Distributed File System (HDFS) that replaces HDFS’ single node in-memory metadata service, with a distributed metadata service built on a NewSQL database.

尤其是在 Spotify 的測試，有 16～37 倍的改善 (應該是指碰到 HDFS 時的這塊，像是從外部拉到 HDFS 上的分析，而非整體的效率改善)：

Metadata capacity has been increased to at least 37 times HDFS’ capacity, and in experiments based on a workload trace from Spotify, we show that HopsFS supports 16 to 37 times the throughput of Apache HDFS.

論文裡面有提到用的是 MySQL Cluster 的 NDB (in-memory)：

HopsFS stores all metadata normalized in a highly available, in-memory, distributed, relational database called Network Database (NDB), a NewSQL storage engine for MySQL Cluster.

這樣應該會讓 Hadoop 的人有改善方向...

Markdown 轉電子書 (以及實體書) 的工具

看到有人用 Markdown 寫電子書：「How I wrote and published my novel using only open source tools」。

電子書的部份主要是透過 Pandoc 轉換，然後用 Dublin Core Metadata Element Set 提供 metadata。

實體書的部份也還是用 Pandoc 轉換成 ODT，而後面還是拿出 LibreOffice 對 page format 調整：

LibreOffice has many page formatting options that are non-obvious but incredibly important. For example, I wanted the first page of each chapter to be a right page, and to exclude the page header. I defined my chapter heading style to have a “page break before” and set a page style of “first page” (which is a right page without a header), which is followed by “right page”, which is followed by “left page”, which goes back to “right page”.