-
Notifications
You must be signed in to change notification settings - Fork 2.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Change index prefix from : to - #1284
Change index prefix from : to - #1284
Conversation
The readers will still read from indices containing prefix "prefix:" to keep backward compability. Signed-off-by: Pavol Loffay <ploffay@redhat.com>
Signed-off-by: Pavol Loffay <ploffay@redhat.com>
Codecov Report
@@ Coverage Diff @@
## master #1284 +/- ##
======================================
Coverage 100% 100%
======================================
Files 161 161
Lines 7206 7210 +4
======================================
+ Hits 7206 7210 +4
Continue to review full report at Codecov.
|
cc @jaegertracing/elasticsearch |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I assume there are will be some additional migration steps people will need to take, e.g. to configure rotations / cleaner to look for different names?
Also, a meta comment: I would much prefer if we didn't have to deal with these separators in the first place. The --es-index-prefix
parameter should be truly the prefix, not the prefix's prefix, i.e. the user should specify "production-"
as the value, and then our code would not have to deal with this :
vs. -
difference at all.
Is it possible to make such change (remove separators from code) and have a reasonable migration procedure, e.g. by creating aliases with the new separator for the old aliases and then simply re-deploying collectors with the updated prefix value?
@pavolloffay I think Yuri has a point about the index prefix should be the full prefix, without adding an extra separator. Maybe we could add a temporary command line flag, e.g. However this would be more disruptive - so the current PR provides the smoothest migration path. |
@yurishkuro I had exactly the same comment in #1238 (comment). I am keen on this approach adding the separator in other projects is annoying. However this will mean a potential breaking change. If we go with plain prefix without separator can we release as 1.9? We want to release on Monday so please comment. What I have done at the moment does not require any migration. The only thing to keep an eye on is to use the older version of If we go with the plain prefix - without adding a separator it will require a migration (If users want to migrate to
|
Another option is to keep this hardcoded separator for now, do a release without serious migration requirements, and then in the next release remove the separator. Next release will be breaking for configuration, but with easy migration since by then old index names would hopefully TTL-out |
@yurishkuro I am not sure if I am following you. Could you please be specific about what you are proposing?
So noop in this release? Or keep the current state of this PR?
So the next release remove the separator and require it to be added by users? |
I am suggesting that we merge your current PR version which changes the hardcoded separator to Then we can make the change for 1.10 where the separator is not hardcoded, and people would need to change their deployment configuration to pass We can also punt on removing the default separator altogether, to reduce churn, maybe reserve it for some future 2.0 release. |
Resolves #1238
Spark dependencies: jaegertracing/spark-dependencies#52
The readers will still read from indices containing prefix "prefix:"
to keep backward compability.
Signed-off-by: Pavol Loffay ploffay@redhat.com