Ingest pipeline best practices #1381
Open
+1,428
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
There are a couple of things I need help with.
|
philippkahr
added
Team:Platform
kilfoyle
added
Team:Obs
|
Thanks a lot for opening this Philipp! I've added the "Team:Obs" label since under the new docs organization that's where the ingest content will land. |
There was a problem hiding this comment.
@philippkahr I started reviewing this PR, but I didn't get very far (yet!). There's a lot of content to get into! I'm just going to post the comments/questions/suggestions I have so far so I can see if I'm on the right track. I can jump back in next week.
Some themes in my early feedback include:
- I see some opportunities to simplify the examples to really emphasize the point you're making in each section.
- It might be helpful to write out in plain language what the example is trying to achieve before jumping into a code snippet. (I provided a couple suggestions below.)
- There are probably opportunities to remove redundant information.
Comment on lines
+9
to
+11
| # Common mistakes | ||
|
|
||
| Here we are not discussing any performance metrics and if one way or the other one is faster, has less heap usage etc. What we are looking for is ease of maintenance and readability. Anybody who knows a bit about ingest pipelines should be able to fix a lot of issues. This section should provide a clear guide to “oh I have written this myself, ah that is the easier way to write it”. |
There was a problem hiding this comment.
What about reframing it so it's more focused on what it is (rather than what it's not)? Maybe something like this?
Suggested change
| # Common mistakes | |
| Here we are not discussing any performance metrics and if one way or the other one is faster, has less heap usage etc. What we are looking for is ease of maintenance and readability. Anybody who knows a bit about ingest pipelines should be able to fix a lot of issues. This section should provide a clear guide to “oh I have written this myself, ah that is the easier way to write it”. | |
| # Create readable and maintainable ingest pipelines | |
| There are many ways to achieve similar results when creating ingest pipelines, which can make maintenance and readability difficult. This guide outlines patterns you can follow to make the maintenance and readability of ingest pipelines easier without sacrificing functionality. | |
| :::{note} | |
| This guide does not provide guidance on optimizing for ingest pipeline performance. | |
| ::: |
|
|
||
| ## if statements | ||
|
|
||
| ### Contains and lots of ORs |
There was a problem hiding this comment.
This just seems like a best practice for writing code in general. There's nothing specific about ingest pipelines (or am I missing something)?
Comment on lines
+35
to
+54
| ### Missing ? and contains operation | ||
|
|
||
| Here is another example, which would fail if `openshift` is not properly set since it is not using `?`, also the `()` are not really doing anything. As well as the unnecessary check of `openshift.origin` and then `openshift.origin.threadId` | ||
|
|
||
| ```painless | ||
| "if": "ctx.openshift.eventPayload != null | ||
| && (ctx.openshift.eventPayload.contains('Start expire sessions')) | ||
| && ctx.openshift.origin != null | ||
| && ctx.openshift.origin.threadId != null | ||
| && (ctx.openshift.origin.threadId.contains('Catalina-utility'))", | ||
| ``` | ||
|
|
||
| This can become this: | ||
|
|
||
| ```painless | ||
| "if": "ctx.openshift?.eventPayload instanceof String | ||
| && ctx.openshift.eventPayload.contains('Start expire sessions') | ||
| && ctx.openshift?.origin?.threadId instanceof String | ||
| && ctx.openshift.origin.threadId.contains('Catalina-utility')", | ||
| ``` |
There was a problem hiding this comment.
also the
()are not really doing anything
I don't think this adds much value. Can we simplify these examples so the difference between before/after are easier to see?
We could end up with something like below instead. Let me know what you think about this approach.
### Null safe operator
Anticipate potential problems with the data, and use the [null safe operator](elasticsearch://reference/scripting-languages/painless/painless-operators-reference.md#null-safe-operator) (`?.`) to prevent data from being processed incorrectly.
For example, if you only want data that has a valid string in a `ctx.openshift.origin.threadId` field:
#### **Don't**: Leave the condition vulnerable to failures and use redundant checks
```painless
ctx.openshift.origin != null <1>
&& ctx.openshift.origin.threadId != null <2>
```
1. It's unnecessary to check both `openshift.origin` and `openshift.origin.threadId`.
2. This will fail if `openshift` is not properly set because it assumes that `ctx.openshift` and `ctx.openshift.origin` both exist.
#### **Do**: Use the null safe operator
```painless
ctx.openshift?.origin?.threadId instanceof String <1>
```
1. Only if there's a `ctx.openshift` and a `ctx.openshift.origin` will it check for a `ctx.openshift.origin.threadId` and make sure it is a string.
Comment on lines
+56
to
+79
| ### Contains operation and null check | ||
|
|
||
| This includes an initial null check, which is not necessary. | ||
|
|
||
| ```painless | ||
| "if": "ctx.event?.action !=null | ||
| && ['bandwidth','spoofed syn flood prevention','dns authentication','tls attack prevention', | ||
| 'tcp syn flood detection','tcp connection limiting','http rate limiting', | ||
| 'block malformed dns traffic','tcp connection reset','udp flood detection', | ||
| 'dns rate limiting','malformed http filtering','icmp flood detection', | ||
| 'dns nxdomain rate limiting','invalid packets'].contains(ctx.event.action)" | ||
| ``` | ||
|
|
||
| This behaves nearly the same: | ||
|
|
||
| ```painless | ||
| "if": "['bandwidth','spoofed syn flood prevention','dns authentication','tls attack prevention', | ||
| 'tcp syn flood detection','tcp connection limiting','http rate limiting', | ||
| 'block malformed dns traffic','tcp connection reset','udp flood detection', | ||
| 'dns rate limiting','malformed http filtering','icmp flood detection', | ||
| 'dns nxdomain rate limiting','invalid packets'].contains(ctx.event?.action)" | ||
| ``` | ||
|
|
||
| The difference is in the execution itself which should not matter since it is Java under the hood and pretty fast as this. In reality what happens is the following when doing the first one with the initial: `ctx.event?.action != null` If action is null, then it will exit here and not even perform the contains operation. In our second example we basically run the contains operation x times, for every item in the array and have `valueOfarray.contains('null')` then. |
There was a problem hiding this comment.
This example confuses me. Why would you want to run the contains operation n times if you already know ctx.event.action is null and it's going to return false.
Comment on lines
+95
to
+115
| Similar to the one above, in addition to that, we do not have `?` for dot-walking. | ||
|
|
||
| ```json | ||
| { | ||
| "fail": { | ||
| "message": "This cannot be parsed as it a list and not a single message", | ||
| "if": "ctx._tmp.leef_kv.labelAbc != null && ctx._tmp.leef_kv.labelAbc instanceof List" | ||
| } | ||
| }, | ||
| ``` | ||
|
|
||
| This version is easier to read and maintain since we remove the unnecessary null check and add dot walking. | ||
|
|
||
| ```json | ||
| { | ||
| "fail": { | ||
| "message": "This cannot be parsed as it a list and not a single message", | ||
| "if": "ctx._tmp?.leef_kv?.labelAbc instanceof List" | ||
| } | ||
| }, | ||
| ``` |
There was a problem hiding this comment.
Do these have to be two separate examples? Why are some examples painless and some json?
Comment on lines
+163
to
+178
| ### Checking null way to often | ||
|
|
||
| This: | ||
|
|
||
| ```painless | ||
| "if": "ctx.process != null && ctx.process.thread != null | ||
| && ctx.process.thread.id != null && (ctx.process.thread.id instanceof String)" | ||
| ``` | ||
|
|
||
| Can become just this: | ||
|
|
||
| ```painless | ||
| "if": "ctx.process?.thread?.id instanceof String" | ||
| ``` | ||
|
|
||
| That is what the `?` is for, instead of listing every step individually and removing the unnecessary `()` as well. |
There was a problem hiding this comment.
It feels like this lesson has already been learned on this page by the time you get here.
Comment on lines
+133
to
+145
| ### Checking null | ||
|
|
||
| It is not necessary to write a `?` after the ctx itself. For first level objects such as `ctx.message`, `ctx.demo` it is enough to write it like this. If ctx is ever null you face other problems (basically the entire context, so the entire `_source` is empty and there is not even a _source... it's basically all null) | ||
|
|
||
| ```painless | ||
| "if": "ctx?.message == null" | ||
| ``` | ||
|
|
||
| Is the same as: | ||
|
|
||
| ```painless | ||
| "if": "ctx.message == null" | ||
| ``` |
There was a problem hiding this comment.
Can this be covered in a note in the second example with something like this?
:::{tip}
It is not necessary to use a null safe operator for first level objects
(for example, use `ctx.openshift` instead of `ctx?.openshift`).
`ctx` will only ever be `null` if the entire `_source` is empty.
:::
Comment on lines
+81
to
+93
| ### Checking null and type unnecessarily | ||
|
|
||
| This is just unnecessary | ||
|
|
||
| ```painless | ||
| "if": "ctx?.openshift?.eventPayload != null && ctx.openshift.eventPayload instanceof String" | ||
| ``` | ||
|
|
||
| Because this is the same. | ||
|
|
||
| ```painless | ||
| "if": "ctx.openshift?.eventPayload instanceof String" | ||
| ``` |
There was a problem hiding this comment.
This is what this section could look like using the same structure as the section above.
### Use null safe operators when checking type
If you're using a null safe operator, it will return the value if it is not `null` so there is no reason to check whether a value is not `null` before checking the type of that value.
For example, if you only want data when the value of the `ctx.openshift.origin.eventPayload` field is a string:
####  **Don't**: Use redundant checks
```painless
ctx?.openshift?.eventPayload != null && ctx.openshift.eventPayload instanceof String
```
####  **Do**: Use the null safe operator with the type check
```painless
ctx.openshift?.eventPayload instanceof String
```
Comment on lines
+117
to
+131
| ### Checking null and for a value | ||
|
|
||
| This is interesting as it misses the `?` and therefore will have a null pointer exception if `event.type` is ever null. | ||
|
|
||
| ```painless | ||
| "if": "ctx.event.type == null || ctx.event.type == '0'" | ||
| ``` | ||
|
|
||
| This needs to become this: | ||
|
|
||
| ```painless | ||
| "if": "ctx.event?.type == null || ctx.event?.type == '0'" | ||
| ``` | ||
|
|
||
| The reason why we need twice the `?` is because we are using an OR operator `||` therefore both parts of the if statement are executed. |
There was a problem hiding this comment.
This is what this section could look like using the same structure as the sections above.
### Use null safe operator with boolean OR operator
When using the [boolean OR operator](elasticsearch://reference/scripting-languages/painless/painless-operators-boolean.md#boolean-or-operator) (`||`), you need to use the null safe operator for both conditions being checked.
For example, if you want to include data when the value of the `ctx.event.type` field is either `null` or `'0'`:
####  **Don't**: Leave the conditions vulnerable to failures
```painless
ctx.event.type == null || ctx.event.type == '0' <1>
```
1. This will fail if `ctx.event` is not properly set because it assumes that `ctx.event` exists. If it fails on the first condition it won't even try the second condition.
####  **Do**: Use the null safe operator in both conditions
```painless
"if": "ctx.event?.type == null || ctx.event?.type == '0'"
```
1. Both conditions will be checked.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
based on the discussions here: #1052
this is my first PR against the docs, and I am building a couple of new pages. I think it makes sense to split it out. I am putting it into that part of the docs. https://www.elastic.co/docs/manage-data/ingest/transform-enrich/ingest-pipelines The tips and tricks are generic and not specific to just o11y, or security.