Multiple nodes overlayed on one page, in new Marvel?

[aaron_ximm]

In the process of migrating from ES 1.x to 2.x, getting used to the new Kibana-based Marvel.

In the 'old' Marvel I had views set up to monitor e.g. all the client nodes in my cluster in one view, all the data nodes in another,etc...

...with the 'new' Marvel for 2.x, I can't figure out a way to 'multiselect' multiple nodes, and then view their graphs at a glance.

This was invaluable as outliers popped out instantly... has this been deprecated(!)...?

I tried mutating the URL directly, (e.g. changing ...app/marvel#/nodes/<ip:port>?... in various ways like to a comma-delimited list, or, using wildcards...

...to no avail.

Am I out of luck?

aaron

[bhatch (Brandon Hatch)]

I discovered this previously. That and a few other pieces have kept us from upgrading to the newest versions of Elasticsearch.

Missing Metrics in Marvel 2.X compared to 1.X?

Has anyone converted from Marvel 1.X to 2.X? We are currently on 1.X and are looking into upgrading so we spun up a small instance with the newest versions of ElasticSearch and Marvel (2.1.1 at the time). We then put in a small bit of test data. However a lot seems to be missing. For one, the documentation for the newest version has less info compared to the old versions. 1.3 had information describing each dashboard and what was on it. 2.2 just has info on how to install it. https://www.elast…

I never got a good answer as to whether they will be bringing it back or not.
Some of the data is still stored in the marvel index, it just isn't displayed anywhere. So you could build your own Kibana visualizations to get some of the same data back. Just not sure if it will work quite as well.

[pickypg (Chris Earle)]

Building the Kibana dashboard has been a workaround that I've seen used, and it should be doable in the short term (and long term for that matter, but it should hopefully become unnecessary).

In the longer term, we are bringing back some (not all because some of them were not actually helpful) of the missing metrics and screens.

[aaron_ximm]

Hi Chris,

Should I take as true then that it is not currently possible to monitor multiple (many) nodes on a single page?

This was a pretty core feature in prior versions of Marvel; flipping between many different pages each doing its own expensive refresh etc. is a completely different UX. :

-aaron

[tsullivan (Tim Sullivan)]

Hi Aaron, it is true that Marvel doesn't give you the ability to monitor multiple nodes on a single page. It is an intentional change in theUX.

flipping between many different pages each doing its own expensive refresh etc. is a completely different UX

While it might not completely satisfy you to hear, a solution I often see recommended is to use a browser plugin such as "Revolver - Tabs" for Chrome to automatically toggle between the screens you want to see the most. My office uses this method to toggle between different Kibana dashboards at an interval, and the same idea could apply to toggling between different detail views of nodes in Marvel.

[aaron_ximm]

Hi Tim,

I'm sure it was intentional in the sense of, not a bug...

...but it is a major degradation of functionality.

Today for example I have spent the better part of the day attempting to discern why multiple nodes in my 2.3.1 cluster a) failed to rotate their logs, and b) as a result stopped logging entirely, which we believe also explains why performance degraded over six hours until the cluster had to be pulled from service.

In the past, it was trivially easy to examine all data nodes across many metrics.

Today, not only do I have to bring my laptop to a stand-still maintaining N open tabs, but also as far as I can tell, 95% of the useful default Marvel metrics are simply missing.

No details on anything... I have only realtime cat to turn to.

Very frustrating day as a result...

Not sure where to turn.