Move user related fields to align with ECS schema. #1783
Conversation
| @@ -63,6 +63,9 @@ | |||
| "started": 43 | |||
| }, | |||
| "type": "request" | |||
| }, | |||
| "user": { | |||
| "id": "123user" | |||
There was a problem hiding this comment.
user information from the metadata is set if no transaction.context.user information was sent.
| @@ -80,7 +80,10 @@ func (m *Metadata) Merge(fields common.MapStr) common.MapStr { | |||
| utility.Add(fields, "host", m.System.fields()) | |||
| utility.Add(fields, "process", m.Process.fields()) | |||
| utility.MergeAdd(fields, "service", m.Service.fields()) | |||
| utility.MergeAdd(fields, "user", m.User.fields()) | |||
| utility.AddIfNil(fields, "user", m.User.Fields()) | |||
There was a problem hiding this comment.
I don't think we should merge user information from metadata and context.user but overwrite it.
There was a problem hiding this comment.
@graphaelli the user_agent and the client.ip are set, even if the context.user information is set directly in an event, as they are nested under different root keys and not under user. So using AddIfNil seems proper here.
There was a problem hiding this comment.
Agreed on both counts. Think we need an integration test for this too?
There was a problem hiding this comment.
The MergeAdd was introduced long time ago, https://github.com/elastic/apm-server/pull/687/files.
I think it is a bug to mix two potential different user informations to one, that is why I changed it to AddIfNil, which basically first checks if there was a user sent with the specific event, and if not sets the one from the metadata.
The possibility to send a user in metadata and the event and if this makes sense or not is a completely different topic to discuss. Please raise any concerns about this in this issue #1255.
Move `context.user.email` to `user.email` Move `context.user.id` to `user.id` Move `context.user.username` to `user.name` Move `context.user.ip` to `client.ip` Move `context.user.user-agent` to `user_agent` fixes elastic#1782
simitt
force-pushed
the
ecs-change-user
branch
from
f0db328
to
dc56f56
Compare
|
Kibana Index PR is up elastic/kibana#28678 |
|
@graphaelli @jalvz is there something left you require for this PR? Otherwise I'd appreciate an approval. |
|
jenkins, retest this please. |
|
jenkins, retest this |
|
jenkins, retest this please |
| @@ -80,7 +80,10 @@ func (m *Metadata) Merge(fields common.MapStr) common.MapStr { | |||
| utility.Add(fields, "host", m.System.fields()) | |||
| utility.Add(fields, "process", m.Process.fields()) | |||
| utility.MergeAdd(fields, "service", m.Service.fields()) | |||
| utility.MergeAdd(fields, "user", m.User.fields()) | |||
| utility.AddIfNil(fields, "user", m.User.Fields()) | |||
There was a problem hiding this comment.
Agreed on both counts. Think we need an integration test for this too?
|
Great work! Really thorough testing 💯 |
|
Merging this in now as the failing test is a flaky, unrelated one. |
Move
context.user.emailtouser.emailMove
context.user.idtouser.idMove
context.user.usernametouser.nameMove
context.user.iptoclient.ipMove
context.user.user-agenttouser_agentfixes #1782
depends on elastic/kibana#28678