[Koha] [Koha-devel] Switching Koha to ElasticSearch... any problems?

Thomas Klausner domm at plix.at
Fri Apr 1 02:15:57 NZDT 2022


Hi!

On Thu, Mar 31, 2022 at 03:52:45PM +1300, Mason James wrote:

> does anyone have any experience running Koha with ES?
> 
> i've done some basic testing (on 20.11) - it seems to work well and its speed is impressive! (🦄 + 🌈)
> 
> - is it generally safe to run in production?
> - any major problems, or missing areas of functionality?

We launched a big library using ElasticSearch this January, it work very 
well, but:

* you need to make sure to give ES the proper resources on the server.
  It will need more RAM than Zebra...

* I discovered (and fixed) a few bugs in the ElasticSearch Mappings 
  Tool, until they are merged (and published in a new release) be very 
  careful with this DataTable! (30142, 29893, 29010)

* Not all searches use those mappings, eg auth_finder.pl uses hardcoded 
  MARC fields

* ElasticSearch interprets lowercase boolean ops as literal search 
  terms. So "foo or bar" finds docs that contain "foo" and "or" and 
  "bar", but "foo OR bar" finds docs containing either "foo" or "bar".
  AFAIK in 20.11 most of these problems are fixed, but see 30153 for a 
  rather recent fix (which I think is already available in the most 
  recent point releases)

* One thing that proved problematic during our initial migration of 500k 
  records was that ElasticSearch indexing is automatically triggered 
  after each edit for each item. This is fine when editing / catalouging 
  by hand, because all changes are immediatly available. But during 
  migration it caused a huge performance hit. There is / was some work 
  to offload the indexing to a worker process, but I think that's not 
  done / stalled. So we hacked our instance to skip the immediate 
  indexing via an ENV var (much easier than passing a command line flag 
  from the script through several layers of modules)

having said that, I think going with ES is a very smart move, because 
(other then Zebra) it's a standard tool with very good docs and a lot of 
knowledgable people (i.e. stackoverflow) to help you :-)

Greetings,
domm

-- 
#!/usr/bin/perl                             https://domm.plix.at
for(ref bless{},just'another'perl'hacker){s-:+-$"-g&&print$_.$/}


More information about the Koha mailing list