Repository navigation
APM latency graph is not filtering properly on 8.8 #158964
Description
Activity
Pinging @elastic/apm-ui (Team:APM)
- added a commit that references this issue
on Jun 6, 2023 Hi @danmcgee-soda, thanks for reporting. I've put up a bug fix at #159085. As a workaround, you can disable service overview metrics (available in Kibana advanced settings) or add the following filter to the search bar of the APM UI:
(NOT metricset.name:service_transaction).Reacted by Dan McGee@insukcho the bug was reported for 8.8.1 (which is what this issue is for, a report), but was fixed for 8.8.2 (the PR itself is only labeled 8.8.2)
@dgieselaar when checking the top of initial thread, it said the issue was reported for 8.8.0.
@insukcho yes you are right :) in general I would ignore the version label on the reported issue. We sometimes use it for planning but it's also sometimes used for the version it was reported for. The label on the PR is ~always right though.
@insukcho we also have an SKI that correctly reports 8.8.0 as the introduced version, and 8.8.2 as the fixed version - is this misconception an internal or public concern?
@dgieselaar I saw some user was confused that the fix was in 8.8.1 while my understanding was aligned with yours. Just want to double check my understanding was correct and tried to get rid of any upcoming confusion for the future reference. Thanks!!
Kibana version:
8.8.0
Elasticsearch version:
8.8.0
Server OS version:
Elastic Cloud
Browser version:
Chrome
Browser OS version:
OS X 13.2.1
Original install method (e.g. download page, yum, from source, etc.):
Elastic Cloud
Describe the bug:
When viewing a single transaction within a service in the APM UI, the latency graph appears to reflect the average/p95/p99 of all transactions, not the selected one. Viewing any transaction within the same service shows an identical graph.
Steps to reproduce:
Expected behavior:
The graph for each transaction should be different, and only include the measured latency for that single transaction, not the average/p95/p99 latency of the entire service.
Screenshots (if relevant):
Transaction "A" - this is a known slow endpoint of ours. Note that the latency spikes shown correspond to the throughput graph, but there are also other much shorter measured latencies that don't show up on the throughput graph.
Transaction "B" - this particular endpoint is only called at a single time period each day. Note that the latency graph is identical, and shows spikes in latency when there was zero traffic, but the throughput graph is very different:
Errors in browser console (if relevant):
N/A
Provide logs and/or server output (if relevant):
N/A
Any additional context:
We upgraded directly from 8.6.2 to 8.8.0, so I can't definitively say that this wasn't broken in 8.7 as well. However, given I cannot find another bug report about this, I would expect it was introduced recently.