cmoulliard
commented
7 years ago Two points should be discusssed/addressed concerning APM:
- Backend : ES vs Apache Cassandra
- UI : OpenShift Web Console vs Grafana Console vs "Obsidian" = Business Activity Console
- Tracing: @Trace, Byteman, ... - how can define APM in an project and next collect the info
Thomas Heute's comment
"Hawkular APM / Distributed tracing is considered a P2. IMO we should still make progress in community, gauge the interest and keep discussing it. I would tend to agree that having it in v1.0 of Obsidian is too premature.
The application metrics (and JVM Metrics) has been mentioned several times in the past, this would IMO be achievable, so far we have issues to work with the OpenShift team, they often have concerns on what we propose as solution (without providing any alternative/viable solution). I'm concerned when John Clingan says we "drop" Hawkular to use Jolokia, since something needs to collect from Jolokia then the data need to be presented somewhere, I'm also hearing that Fabric8/hawt.io might be the solution at least in the interim ?
This is where I get confused, because the instructions to focus on app metrics and JVM comes from Joe Fernandes and Diogenes, and Mark told me that Joe gets to decide, this is why APM was pushed down as P2.
Don't get me wrong, I'd love to ship APM, we get quite some interest. We also still need to fix the storage decision. It's using ElasticSearch at the moment (same as ECL), but I'm also hearing concerns there. If we prefer to move to Cassandra for support reasons, we could do it, I've actually told the team to be prepared to switch or support both, but at this time, we try to include features that are in the interesting, the ElasticSearch usage is an inheritance from RTGov but it does the job so far."