Skip to content

Latest commit

 

History

History
89 lines (53 loc) · 3.9 KB

2022-03-03-zammad-elasticsearch-not-running.md

File metadata and controls

89 lines (53 loc) · 3.9 KB
Raw

Elasticsearch not running on Zammad

Date: 2022-03-03

Problem

I wanted to Set up ES for Zammad

After having access to the machine (which up to now is gracefully hosted by CQuest), I verify, as told by CQuest, that elasticsearch was installed.

The problem was the elasticsearch service was not running, according to systemctl status elasticsearch.

journalctl -r -u elasticsearch
shows this error:
Failed to start elasticsearch due to a fatal signal received by control process (code=killed, signal=9/KILL)

Solution

After searching about in various directions, it reminds me of a OOMKill situation. Although we didn't have a OOMKill message in /var/log/syslog because this is an LXC container, so OOMKill happens in the guest.

As I look into /etc/elasticsearch/jvm.options I've seen that by default it handles memory alone by looking at available memory.

Being in an LXC container, it sees more memory than the container is allowed to take. Hence the kill situation.

Thus I decided to fix the heap size manually. I created a file /etc/elasticsearch/jvm.options.d/memory.options 1 with:

-Xms300m
-Xmx300m

After a systemctl start elasticsearch, elasticsearch was up.

Resolution 2 - zammad messed up

I first had a hardtime figuring out how zammad was running.

The real services in /etc/systemd/system/ are the one with a -1 in their name (others are just sleep commands, they act as placeholders to handle dependencies, I guess). But the executable was not there !

This was because I unadvertantly de-installed it, while trying to solve elasticsearch issue (where I did a uninstall / reinstall of elasticsearch). I did not see it was also de-installing zammad :-/

Hopefully I did not purge, and an apt install zammad solved the issue !

Finish ES integration

Following https://docs.zammad.org/en/latest/install/elasticsearch.html, I wanted to verify settings:

$ zammad run rails r 'print Setting.get("es_index") + "\n"'
zammad

I reindex everything using
zammad run rake searchindex:rebuild
as indicated on zammad issue 1630

It failed with:

rake aborted!
Unable to send Ticket.find(33).search_index_update_backend backend: #<RuntimeError: Unable to process post request to elasticsearch URL 'http://localhost:9200/zammad_production_ticket/_doc/33?pipeline=zammad116611039594'. Elasticsearch is not reachable, probably because it's not running or even installed.

Response:
#<UserAgent::Result:0x0000563cf7686228 @success=false, @body=nil, @data=nil, @code=0, @content_type=nil, @error="#<Errno::ECONNREFUSED: Failed to open TCP connection to localhost:9200 (Connection refused - connect(2) for \"localhost\" port 9200)>", @header=nil>

Indeed systemctl status elasticsearch shows me it failed. Due to Terminating due to java.lang.OutOfMemoryError: Java heap space

I edit /etc/elasticsearch/jvm.options.d/memory.options to give more memory:

-Xms600m
-Xmx600m

Then relaunched:

zammad run rake searchindex:rebuild

Finally it works !

Lessons learned

This Elasticsearch memory problem is a recurring one ! In this case it was hard to spot at first sight. But this is always a lead to investigate on a hard kill of an Elasticsearch instance.

Be prudent when doing apt remove to verify you aren't removing something else ! 😓

Footnotes

  1. in fact I was induced in error because in jvm.options the mandatory .options extension was not cited, so I first created a file with no extension and it did not work (but no log could tell me that). Reading online doc I sort this out. The problem is already fixed in elasticsearch repos ↩