You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add an optional monitoring of topic processing delay to ingester pipeline (#4453)
This will:
- Expose metrics to show how far behind each partition of the topic
being consume is in ms
- We utilise peekMessages on the pulsar REST/admin api to get this
information
- It is optional and off by default
The purpose of this is so you can see how far behind the ingester is in
time (i.e 3 minutes)
- pulsar only exposes backlog in number of messages, so its hard to tell
how much impact a delay is having (i.e 1000 messages could be 1 second
behind or 10 minutes)
Also exposes a metric that shows the publish time of messages being
processed by the pipeline for each partition
**Note** I've had to "fork" some of pulsar-client-go due to a bug in the
client. This is being fixed in
apache/pulsar-client-go#1419. Once that is in
the upstream client we can remove this internal code copy/fork
---------
Signed-off-by: JamesMurkin <jamesmurkin@hotmail.com>
// This file is largely a copy of https://github.com/apache/pulsar-client-go/blob/230d11b82ba8b60c971013516c4922afea4a022d/pulsaradmin/pkg/admin/admin.go#L1
2
+
// With simplifications to only cover the functionality we need
3
+
// This is to work around a bug in the client which is tracked here:
0 commit comments