Twelve Factor: Difference between revisions
From ym2149.org
Jump to navigationJump to search
No edit summary |
No edit summary |
||
| (4 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
<div style="float: right"> | |||
__TOC__ | |||
</div> | |||
* https://12factor.net/ | * https://12factor.net/ | ||
* itil for influencers | * itil for influencers | ||
| Line 8: | Line 11: | ||
== V == | == V == | ||
* a deployment should only be able to access config specific to that environment | * as insurance against unpleasant surprises, a deployment should only be able to access config specific to that environment (along with config common to all envs) | ||
== X == | == X == | ||
| Line 16: | Line 19: | ||
== XI == | == XI == | ||
* logging goes to stderr not stdout | * logging goes to stderr which is already unbuffered and this is consistent with the rest of the world, not stdout which is fully buffered (would be line buffered if connected to a terminal) | ||
* wishful thinking to characterise logging as an event stream as much of it is under the control of libs or even other processes, and logs are typically untested | |||
* structured logging to satisfy the demands of visual tools is expensive, plain text is sufficient for investigating incidents using ordinary unix tools, and it compresses well | |||
[[Category:Programming]] | [[Category:Programming]] | ||
Latest revision as of 19:47, 6 November 2024
- https://12factor.net/
- itil for influencers
III
- environment variables are too crude for typical config workloads
V
- as insurance against unpleasant surprises, a deployment should only be able to access config specific to that environment (along with config common to all envs)
X
- differences between envs must be inspectable, so that the confidence you gain in one env actually means something in the next
XI
- logging goes to stderr which is already unbuffered and this is consistent with the rest of the world, not stdout which is fully buffered (would be line buffered if connected to a terminal)
- wishful thinking to characterise logging as an event stream as much of it is under the control of libs or even other processes, and logs are typically untested
- structured logging to satisfy the demands of visual tools is expensive, plain text is sufficient for investigating incidents using ordinary unix tools, and it compresses well
