[Failure store] Remove ::* selector
#121900
Conversation
elasticsearchmachine
added
the
needs:triage
gmarouli
added
>non-issue
:Data Management/Data streams
|
Pinging @elastic/es-data-management (Team:Data Management) |
| public static IndexComponentSelector read(StreamInput in) throws IOException { | ||
| return getById(in.readByte()); | ||
| byte id = in.readByte(); | ||
| if (in.getTransportVersion().onOrAfter(TransportVersions.REMOVE_ALL_APPLICABLE_SELECTOR)) { |
There was a problem hiding this comment.
More versions will be added here when it gets backported.
server/src/main/java/org/elasticsearch/cluster/metadata/IndexNameExpressionResolver.java
Outdated
Show resolved
Hide resolved
| DATA("data", (byte) 0), | ||
| FAILURES("failures", (byte) 1), | ||
| ALL_APPLICABLE("*", (byte) 2); | ||
| FAILURES("failures", (byte) 1); |
There was a problem hiding this comment.
This is an unlikely scenario but could we ever end up in a place where a new selector added here could collide with the old ::* value in a mixed cluster?
There was a problem hiding this comment.
Hm.... Interesting thought. Thinking out loud because this is a tough cookie.
If we implement this the way it's now and backport it back to 8.18 and 8.x.
We can remove this backwards compatibility code from 9.x; because you always have to go through the 8.last version. 8.last will never propagate this selector because it matches it to ::data and this selector is never persisted anywhere, I think. Furthermore, there is no CCR or CCS supported yet. Right?
This means that if we ever introduced ::* in 9+ we could distinguish the two. Right?
3 participants
To simplify the failure store API we remove the wildcard selector
::*.When a user wants to access both data and failure components they can write
my-ds,my-ds::failuresormy-ds::data,my-ds::failures.This should simplify the security model as well.