Elasticsearch died while starting up

Elastic Stack Elasticsearch
1

Hi All,
Running into this error while I am trying to deploy Elasticsearch v8.17.1 on ppc64le machine
Gradle Version : 8.13

OS Info : Linux 5.14.0-503.38.1.el9_5.ppc64le (ppc64le)

JDK Version : 21.0.6+7-LTS (Eclipse Temurin)

JAVA_HOME : /opt/adopt

{"type": "server", "timestamp": "2025-05-19T07:26:13,647Z", "level": "INFO", "component": "o.e.n.NativeAccess", "cluster.name": "docker-cluster", "node.name": "34fe09bae17b", "message": "Using [jdk] native provider and native methods for [Linux]" }

ERROR: Elasticsearch died while starting up, with exit code 92
2

If you look at the official support matrix you can see that ppc64le is not a supported architecture. I would recommend deploying it on a different machine with a supported architecture.

I recall other users trying to run Elasticsearch on unsupported hardware architectures in the past and if I remember correctly they had a lot of problems and were not able to get much support.

3

@christian, we have rebuilt the binary on ppc64le and during runtime we get this exit 92 related issue...Issue does not seems to be arch specific..what are all the possible reason for "exit code 92" ? we checked ulimit -l is unlimited and file desc are 65k...anything more to investigate ? details around exit 92 helps..thanks..

4

In recent versions I would set this a lot higher. Try that and see if that helps.

Apart from that suggestion I am not able to help.

5

Do you mean that you can reproduce the issue on a supported architecture? If so, please share more details.

2 Likes
6

Hello,

Could you please see if there is anything extra in the elasticsearch log file in order to see if it can help to find the root cause?

Thanks!!

7

@Tortoise do you mean the log4j2.properties file?

8

Hello,

Not the properties file but the elasticsearch log file ( elasticsearch.log) which will have more details about the error when you try to start the elasticsearch service but it is erroring out with below information
ERROR: Elasticsearch died while starting up, with exit code 92

Thanks!!

9

@Tortoise that's all we have in the log file

1 Like
10

Run within a debugger. Use strace or similar tools. Check the source.

Nothing wrong with being curious and setting yourself a little challenge. But IF you get it to work, you’re not going to do anything important with it, right? There’s no business involved? It’s a hobby project, right?

11

Thanks for the debug tips. We listed out the below options for exit 92 failure (i) Invalid JVM options (ii) JVM not found (iii) memory locking (iv) resource limitation (v) seccomp filter..as of now, ruled out the resource limitations by setting files, locked memory to unlimited..also, JVM related issue also kind of ruled out since JVM is coming up and it fails with "ServerProcessBuilder.start" within Elasticsearch server...further investigation is on..it is not a hobby project as we are looking forward to use ESv8 in our product which will be deployed on Power architecture

12

Er .... thats very brave of you.

The

advice should be heeded. IMHO.

You ask on the forum for advice, you get advice. Up to you if you follow it or not.

13

As stated earlier I would recommend against this. As this is a very unusual and unsupported approach you are likely to be on your own if you face any issues. I doubt there are many here that are able to help out.

If you do run into further issues and decide to post here, please make sure that you point out that you are deploying a custom solution on an unsupported architecture as that will save everyone a lot of time and effort.

1 Like
14

what is the process to make it a supported approach ? I created an issue but it got closed.Thanks in advance for your advice.

15

As far as I know there is none. For this to happen Elastic as a company would need to decide to make it a supported architecture/platform and consider this during development and also test on it. This requires a huge investment and as I would consider it a niche architecture I think this is very, very, very unlikely to happen.

16

Echoing what Christian says, the problem is simply down to money. Adding a new supported architecture is a very long-term investment with rather high up-front costs, and with so few users of this CPU architecture right now the business case simply doesn't hold up to scrutiny.

Edit to add: this is something we actively monitor - if the economics of the situation were to change then we'd react accordingly.