-
Notifications
You must be signed in to change notification settings - Fork 490
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
make metrics from carbon-cache go via relay config #959
base: master
Are you sure you want to change the base?
Conversation
This obviously isn't an ideal patch, as it requires a valid relay config to exist in the cache. Our solution is to set up the relay to submit to the HA router port, hence the Probably this should be a config option, something like |
82fcf32
to
fbafb9a
Compare
Update just splits out the |
@bucko909 : tests should be amended for changed behaviour, could you please amend it? Thanks! |
I can do, but I think I need agreement on what the changed behaviour should be -- what's here cannot be shipped as it'll break existing configs. "Proper" approaches are:
I'll submit a patch with the former behaviour and a warning for now; you can decide if that's appropriate. The tests will then work fine if I do this, though you may also want to require a test for the new behaviour. I'll see if I can get one to work. |
04bce00
to
e50d01a
Compare
Codecov ReportAttention: Patch coverage is
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #959 +/- ##
==========================================
+ Coverage 50.63% 50.67% +0.03%
==========================================
Files 36 36
Lines 3446 3467 +21
Branches 535 528 -7
==========================================
+ Hits 1745 1757 +12
- Misses 1574 1582 +8
- Partials 127 128 +1 ☔ View full report in Codecov by Sentry. |
I've added the setting and some warnings in the config file. I'm not sure it'd be possible to add a warning about configuration easily, because one daemon needs to warn depending on another's config, which isn't really in the design spec of the configuration's loader. Some caveats:
Let me know if you think this is reasonable. |
This is necessary for any hashed sharding setup: The cache will write metrics to its local cache, not to the correct cache for the hashing scheme. This means only datapoints synced to disk are available. Worse, in a highly available setup, metrics differ depending on which node you happen to be fetching from.