Hi!

We install Koha in a cluster of LXC containers:

One haproxy for dns redirection and ssl offloading
One container for MariaDB
One for OPAC
One for staff client
One for Zebra
One for misc stuff, like the SIP2-server
One for ElasticSearch
One for Memcached
+containers for other support services

With haproxy it is rather straight forward to start load balancing multiple "web bricks" if you choose to. We don't need to do it since we chose to buy a lot of hardware. Hardware is less expensive than work where we live in.
This clusterization is simply to increase robustness and security.
And actually has saved our bacon few times during the past years, when we get unidentifiable hard-drive freezes with the SIP2-container,
or some cronjob has went awol and completely hogged all system resources available to it.
Without containerization we would have had downtime.
Also helps with simple DOS-attacks which can happen by accident.

I recommend this type of setup.

Unfortunately the debian packages add a lot of unnecesary cruft to each container so we use dev-install instead.

Getting Zebra to work in a separate container:
Zebra is really really really tightly wound into the make-toolchain Koha uses and there is no convenient way of separating it. Maybe you could just bind mount the zebradb-directory between containers. This might be enough to run Zebra separate of the Koha's source code.

Since mass storage is very cheap, we just run two identical Koha instances from the same source code repo. We make several identical Koha-installations via the dev-install.
The Zebra-one only has Zebra and Koha sources (no Apache2, MariaDB, etc packages needed by a standalone Koha).
You can configure Zebra to listen on tcp socket via the $KOHA_CONF. You can conveniently centrally configure the same $KOHA_CONF to all the Koha containers, so they all find the same Zebra by the ip and the same MariaDB. This is best done with Ansible (since community has already started working towards Ansible)


Separating Zebra to a isolated (and easily monitorable) environment is entirely possible.



Here is a snippet from our Ansible infrastructure definitions where Koha is set up with capability: zebra

These should be most of the Ansible steps needed together with Koha-source to get Zebra running standalone:



ansible@hephaestus:~/KSAnsible$ cat roles/koha/tasks/capabilities/zebra.yml

---

- name: Install zebrasrv and zebraidx
  apt:
    name: idzebra-2.0
  become: yes

- name: Create the zebra socket-dir and permissions
  file:
    path: "{{zebra_run_dir}}"
    owner: koha
    group: koha
    state: directory
  become: yes

- name: Create the zebra lock-dir and permissions
  file:
    path: "{{zebra_lock_dir}}"
    owner: koha
    group: koha
    state: directory
  become: yes

- name: Create the zebra data-dir and permissions
  file:
    path: "{{zebra_data_dir}}"
    owner: koha
    group: koha
    state: directory
  become: yes


- name: Configure Zebra stripes
  lineinfile:
    path:   "{{item.path}}"
    regexp: "{{item.regexp}}"
    line:   "{{item.line}}"
  with_items:
    - path:   "{{koha_dev_path}}/etc/zebradb/zebra-biblios-dom.cfg"
      regexp: "^memMax:"
      line:    "memMax: {{zebra_index_mem}}"
    - path:   "{{koha_dev_path}}/etc/zebradb/zebra-biblios-dom.cfg"
      regexp: "^register:"
      line:    "register: {{zebra_data_dir}}/biblios/register:{{zebra_index_size}}"
    - path:   "{{koha_dev_path}}/etc/zebradb/zebra-biblios-dom.cfg"
      regexp: "^shadow:"
      line:    "shadow: {{zebra_data_dir}}/biblios/shadow:{{zebra_index_size}}"
    - path:   "{{koha_dev_path}}/etc/zebradb/zebra-biblios-dom.cfg"
      regexp: "^sortmax:"
      line:    "sortmax: {{zebra_sort_max}}"
    - path:   "{{koha_dev_path}}/etc/zebradb/zebra-biblios-dom.cfg"
      regexp: "^facetNumRecs:"
      line:    "facetNumRecs:{{zebra_facet_num_records}}"
      owner: koha
  become: yes
  notify: Restart Zebra

- name: Zebra - Link Zebra-service
  file:
    src:   "{{koha_dev_path}}/bin/koha-zebra-ctl.sh"
    dest:  "/etc/init.d/koha-zebra-daemon"
    owner: root
    group: root
    state: "link"
  become: yes

- name: Zebra - Enable Zebra-service
  systemd:
    name: koha-zebra-daemon
    enabled: yes
    daemon_reload: yes
    state: started
  become: yes

- name: Zebra-index - Link Zebra-index-service
  file:
    src:   "{{koha_dev_path}}/bin/koha-index-daemon-ctl.sh"
    dest:  "/etc/init.d/koha-index-daemon"
    owner: root
    group: root
    state: "link"
  become: yes

- name: Zebra-index - Enable Zebra-index-service
  systemd:
    name: koha-index-daemon
    enabled: yes
    daemon_reload: yes
    state: started
  become: yes

- name: Deploy Zebra cronjobs
  template:
    owner: "{{koha_user}}"
    src:   "etc_cron.d_koha.j2"
    dest:  "/etc/cron.d/koha-zebra"
  vars:
    cronjobs:
      rbza:
        comment: '#Trigger full Zebra reindexing daily'
        prefix:  ' '
        timing:  '46 10 * * *'
        user:    'koha'
        command: 'migration_tools/rebuild_zebra.pl -b -a -r -v -x'
  become: yes