rob05c commented on a change in pull request #4628: Add ORT Rewrite Blueprint URL: https://github.com/apache/trafficcontrol/pull/4628#discussion_r406978038
########## File path: blueprints/ort-rewrite-unix-style.md ########## @@ -0,0 +1,210 @@ +<!-- +Licensed to the Apache Software Foundation (ASF) under one +or more contributor license agreements. See the NOTICE file +distributed with this work for additional information +regarding copyright ownership. The ASF licenses this file +to you under the Apache License, Version 2.0 (the +"License"); you may not use this file except in compliance +with the License. You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + +Unless required by applicable law or agreed to in writing, +software distributed under the License is distributed on an +"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +KIND, either express or implied. See the License for the +specific language governing permissions and limitations +under the License. +--> +# ORT Rewrite in UNIX Philosophy + +## Problem Description +ORT is: +- Difficult to maintain. Writing Perl is difficult, and reading it is even more difficult. +- Dangerous to modify. Perl is not compiled, and even validity checks (`perl –c`) fail to verify dynamic runtime errors. This makes it very easy to introduce a bug in seldom-executed areas. +- Untested. Perl ORT has no unit or integration tests. +- Opaque. Nobody really knows everything it does, or when, or why. + +## Proposed Change + +ORT will be rewritten into a series of standalone executables, in the "UNIX Philosophy" + +> 1. Make each program do one thing well. To do a new job, build afresh rather than complicate old programs by adding new "features". +> 2. Expect the output of every program to become the input to another, as yet unknown, program. Don't clutter output with extraneous information. Avoid stringently columnar or binary input formats. Don't insist on interactive input. + +- Each executable should do exactly 1 thing, and if a new "thing" becomes necessary, a new executable will be created. +- The input and output of executables should be text which is easily parseable, so the executables can easily be pipelined (passing the output of one to the input of another), as well as easily read by humans and manipulated by standard Linux/POSIX tools. + +This makes ORT: +- Easier to maintain. Each binary does one thing, is much smaller, and is more obvious. Presumably they’re also written in a language easier to read and write, such as Go. +- Safer to modify. If each component is smaller, it’s more obvious what it does. We also presume the apps will be written with good development practices (such as modularization), with a language which verifies more at compile-time, and with tests. +- Clear and easy for operators to understand what each app does. We assume clean interfaces, and good documentation (ideally in the app itself, via help flags, printing usage when no arguments are received, and/or man pages). + +#### Implementation + +The implementation should adhere to the "UNIX Philosophy," POSIX, Linux Standard Base (LSB), and GNU as much as possible. + +ORT will continue to consist of a single OS package (e.g. RPM), which installs all executables. + +ORT will require the following executables: +- **Aggregator**. This is the “primary application” which will emulate the existing ORT script, and be called by CRON or operators to deploy all configs, as ORT does today. Note this is similar to how git works, and several other common Linux CLI utilities. + This app will have no logic itself, except to call the other executables. + - INPUT: configuration and specification to fetch and emplace config files. + - BEHAVIOR: fetches and places config files + - OUTPUT: success or failure message + +- **Traffic Ops Requestor**. This will fetch data needed from Traffic Ops, such as the Update Pending flag, packages, etc. This should never modify TO data, and should be guaranteed read-only. Any status modifications should go in the Traffic Ops Updater. + - INPUT: Traffic Ops URL and credentials, and data to fetch + - BEHAVIOR: Requests data from Traffic Ops + - OUTPUT: Traffic Ops data requested + - Format is probably multipart/mixed, but format may be different if a better format is determined. Ideal "UNIX Philosophy" format is line-delimited text, but the complexity may preclude that. The more complex and difficult to parse, the further from the "UNIX Philosophy." E.g. multipart/mixed is preferable to JSON. +- **Config File Generator**. This will take TO data and produce config files. + - INPUT: Traffic Ops data, and config file(s) to generate or exclude (typically all, possibly “reval only” or other behaviors of ORT) + - BEHAVIOR: No side effects. Computationally: builds requested files. + - OUTPUT: Config files + - Format is probably multipart/mixed, but as above, may be different, multipart is preferable to JSON, etc. +- **Config File Preprocessor**. Preprocesses generated config files, making post-generation modifications such as underscore directive replacements, and remap OVERRIDE replacements. TODO: determine if this should be rolled into the Config File Generator. + - INPUT: Config files, TO data/metadata. + - BEHAVIOR: No side effects; computationally: modifies files per rules and data. + - OUTPUT: modified config files. +- **Server Config Readiness Verifier**. Verifies the operating system is ready and safe to apply the given config. Currently, this is just checking udev rules and verifying ATS block devices don’t have filessytems. But it may be more in the future. TODO: determine if necessary; should this even be ORT’s job? Will we ever need anything besides udev/50-ats.rules? + - INPUT: config files + - BEHAVIOR: No side effects; reads configs and inspects server state. + - OUTPUT: whether server is safe and ready to apply config files. +- **ATS Plugin Readiness Verifier**. Verifies ATS has the necessary plugins for the config files. + - INPUT: config files + - BEHAVIOR: No side effects; reads configs and inspects server state. + - OUTPUT: whether ATS has all necessary plugins, or which files require which missing plugins. +- **Diff Tool**. This will take two config files (presumably an existing and new file) and return their differences. TODO: determine if this is necessary; if configs are deterministic, can POSIX diff be used? + - INPUT: Two config files + - BEHAVIOR: No side effects. Computationally: diffs given files. + - OUTPUT: file diff. Ideally in a standard format. +- **Backup Tool**. This will take a file and copy it to a backup location. TODO: determine if necessary; are standard POSIX cp/mv/etc enough? + - INPUT: config file to backup, and backup location. May be text of new file or path of existing file. + - BEHAVIOR: Save or copy given file to given backup location. Should be atomic. + - OUTPUT: success or failure message. Review comment: >Maybe a backup file name I assume the backup filename is input to this app, and the Aggregator can pick that. The Aggregator could even read a TO Parameter (the answer to all problems) if we wanted. Though that feature doesn't exist today. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
