Working with elasticsearch cluster

Elastic Stack Elasticsearch
1

Hi,
we use elasticsearch for storing some metadata of CI, we have automation's to collect metadata and push to elasticsearch.
Our elasticsearch cluster has below configuration,
2 - master nodes
3 - data nodes
all nodes are individual VM's with different IP's connected into 1 cluster
We configure one of the active master node IP in metadata collectors to push data to ES,
But our question is what if that node goes down for some reason ? We don't want to reconfigure our collectors with new master node IP.
Is there any best way to handle such scenarios ? I guess its quite common scenario when working with clusters

Thanks in advance.

2

A few important things;

Does your collector not support load balancing with multiple hosts?

3

Thanks for correcting :slight_smile:

as of now no, collectors doesn't support load balancing or multiple hosts configuration.
We are looking if this can be done in some other way than collector code change (this we want to handle as a last option)

4

What is the collector, can you share more details on that?

5

Why not add a load balancer between your collector and the cluster?

6

Collects data from our various CI sources,
One is maven plugin: to collect data of Maven artifacts generated in maven build.
Jenkins plugin: to collect non-maven CI data.

7

Right, but does it do that by collecting logs, or something else?

8

We tried it, same question arises there also, no ?
Like what if that LB is down (correct me if i am not wrong , it is also single point of failure)

9

Then I think a code change in the collector may be required. How does your collector handle redundancy?

10

No, not by parsing by logs.
they have different logic, example maven plugin uses maven-core API's and loop through all the projects built and generate report.

11

ok.
for redundancy it follows different way, it writes data into a file and upload to binary repo. When ES is UP later, We have automation which reads these files and push data to ES.

12

Doesn't that make your collector a single-point-of-failure?

13

No, because binary repo is always available.

14

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.