[CELEBORN-688] Add JVM metrics grafana template #1964
Conversation
Codecov Report
@@ Coverage Diff @@
## main #1964 +/- ##
==========================================
- Coverage 47.10% 47.00% -0.09%
==========================================
Files 164 164
Lines 10361 10369 +8
Branches 956 957 +1
==========================================
- Hits 4879 4873 -6
- Misses 5166 5178 +12
- Partials 316 318 +2 see 6 files with indirect coverage changes 📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
| "uid": "_7trpcMIk", | ||
| "version": 5, | ||
| "weekStart": "" | ||
| } |
There was a problem hiding this comment.
Add a new blank line in the end?
|
I want to know if the metrics name will be different in different GC? |
No, It will only change related metrics labels. For specific details, you can refer to #1939. When use ParNew + CMS, the JVM Memory Pools / GC rows look like below: |
|
How about G1? I remember when I use G1 in a test cluster, it was not same in origin code |
You can find the example graphs with g1 in the top PR proposal. Different collectors only have different labels in the new jvm metrics. For pools metrics with different collectors, it can be generated dynamically by the benefit of grafana's replicate panels function(Repeat this panel for each value in the selected variable). |
|
Can you provide a screanshot when it shows the whole cluster? Or it only support show one worker? |
This template now just supports one instance. For JVM metrics, one panel often contains serveal metrics so that we can easily obtain information by comparing related metrics. The cluster perspective may make the diagram a mess. Users can customize the jvm metrics panel of the entire cluster they care about or use explore. But I will change the variable to be multi-selected later to make the template easier to use. |
|
Yea, for me, I always check the whole cluster's status to see if the cluster is healthy. |
Hi, @AngersZhuuuu, I've supported selecting multi/all instances, snapshot like below, please take a look again if have time. |
I have called a increase function on it. https://github.com/apache/incubator-celeborn/pull/1964/files#diff-68dcac07da1db4f5aa934eb21ef636d9fc903f02e85db78acb9f59988baaed0eR946 |
Thanks, merging to main/0.3 |
### What changes were proposed in this pull request? Currently there is no JVM metrics grafana template, nor in grafana labs. For better use, it is necessary to add one. According the change in #1939 This template uses two variables(instance, pool). The layout is divided into 5 rows.  The panels with g1 look like below：  JVM Memory Pools row uses replicated panel mode which panels are automatically deplicated by `pool` variables.     ### Why are the changes needed? Ditto ### Does this PR introduce _any_ user-facing change? Yes, this dashboard is based on changes in #1939 ### How was this patch tested? Cluster test Closes #1964 from onebox-li/add-jvm-dashboard. Authored-by: onebox-li <[email protected]> Signed-off-by: zky.zhoukeyong <[email protected]> (cherry picked from commit 2b79692) Signed-off-by: zky.zhoukeyong <[email protected]>
What changes were proposed in this pull request?
Currently there is no JVM metrics grafana template, nor in grafana labs. For better use, it is necessary to add one.
According the change in #1939
This template uses two variables(instance, pool).
The layout is divided into 5 rows.
The panels with g1 look like below：
JVM Memory Pools row uses replicated panel mode which panels are automatically deplicated by
poolvariables.Why are the changes needed?
Ditto
Does this PR introduce any user-facing change?
Yes, this dashboard is based on changes in #1939
How was this patch tested?
Cluster test