Fix Origin UI RateGraph data loss/issues #497
Conversation
|
@CannonLock - can you take a look at this? |
web_ui/frontend/app/origin/page.tsx
Outdated
| rate={TimeDuration.fromString("3h")} | ||
| duration={TimeDuration.fromString("7d")} | ||
| resolution={TimeDuration.fromString("3h")} | ||
| metric={['xrootd_server_bytes{direction="rx"}', 'xrootd_server_bytes{direction="rx"}']} |
There was a problem hiding this comment.
Lets keep these defaults. I realize they aren't the best when you first enter but for a origin that is actually in use somewhere I think they are more helpful then just todays numbers.
If you disagree, we should at least kick up the resolution a little bit, maybe to "20m"? 1440 data points on the graph is going to get busy.
There was a problem hiding this comment.
I agree we should think about users who run origin for days. How about I add a note to the graph to tell the user if they just started the origin and they can't see incoming data, they should modify the resolution?
There was a problem hiding this comment.
Actually, I can't run origin UI with latest main and I got the white screen with client side error that leads to Objects are not valid as a React child (found: object with keys {Type, Value}). If you meant to render a collection of children, use an array instead. I've rolled back the values to default but I won't be able to add the note without seeing the UI. We can talk about more in detail tomorrow.
|
Given the limited time we have for 7.4, I will back port this in 7.4.x, so no worries if we need to get a release out soon @joereuss12 |
|
@CannonLock I added some instructions in the "no data" error message for how to adjust settings to get data out for fresh servers and I tested everything looks good. |
|
@haoming29 When I run the origin I am getting |
Did you login when you hit this? Where did this come from? |
Fix #487
It turned out that issue #487 is expected as we set the default data resolution to 3h, and at the server fresh start (no history metric data), Prometheus doesn't have any data at
now() - 3h. However, this should be adjust so that the default resolution can always give some result to the user at the server start.This PR fixes anther issue where the "reset" button sets
timedifferently than the default value, causing inconsistent result (although in the later commits it was consolidated to update time accordingly, see details in next point)This PR also fixes a bug where the RateGraph visualizes data from
t= now()tot= now() + range (by default 3h)which means it "generates" fake data from now to 3 hours in the future and we don't want that. Instead, now the data will be fromt=now() - rangetot=now(), which is the history data from last<range>until now.This PR finally addressed the issue where we use
xrootd_server_bytes{direction="rx"}metric for bothrxandtxdata. Now the two data series will give different values (as reported by Prometheus)