Wednesday, February 15, 2023

XMCloud SxaStarter local setup - solr-init issue

 XMCloud - solr-init error on a local container setup with SxaStarter 


Headless SXA

I wanted to try the new headless SXA. As this can be installed on Sitecore 10.3 I installed that version locally and installed SXA on top (tip: a good blog post from Dan about this setup). That went fine, but I bumped into issues with SXA. I asked on StackExchange and Slack but nobody seemed to know the answer (if you do, please answer the question on SSE). I was going to open a support ticket but as this was just a test... well, you know... I'll do that when I need it in a project.

I heard there were differences between SXA in Sitecore 10.3 and the version in XM Cloud. So I decided to try that one.

XM Cloud

As XM Cloud is sort of a SAAS solution there is some discussion about the fact why people would install it locally and it does make sense not to do is - but I was in test and mess-around mode so let's do that on a local setup (just because we can).

As I'm no expert (yet) in containers nor nextJS and such, I was looking for a simple way to do this. With the information I gathered during various sessions about XM Cloud, that should be possible.

The setup 

I found a blog post from Serge van den Oever who already succeeded in such a setup and documented it very well - thanks Serge for the very informative blog.

The setup seemed to go very well...  it takes a while to download all the images but at the end, I had a site that started and I could create the SXA site. Even the creation of the rendering container went fine - some issues mentioned in Serges post seem to be gone, some are still there. But I ended up with a running environment. 

Even Headless SXA worked. So the issues I had on my local XM were not present in this XM Cloud version. Hooray, cheers, all happy...   so why am I writing this post?

The issue

Of course, there had to be an issue. The next day the containers wouldn't start anymore. The solr was going crazy and that means nothing works.  Well in fact, the solr-init won't run because the solr isn't healthy. 

I started two tracks: as I had another laptop I tried the same installation there.. and at the same time started searching on Sitecore StackExchange, Google...    The second installation on the other laptop worked. Once 😞  The result was the same: solr-init would fail when trying to restart the containers. 

In the meantime my search quest led to a blog post from Jeremy Davis that actually didn't sound very hopeful - as you can read in his blog he had the same issue with Sitecore containers (not XMCloud) and already tried several things. So I was not going to try all of those again, but on my machine his workaround "docker network prune" didn't work.

But he does also mention an alternative provided by Rob Ahnemann. So that is a third Sitecore MVP bumping into the same issue 😨  Rob notes that the issue could be solved by removing the Zookeeper - so actually using Solr standalone instead of SolrCloud. This sounds reasonable, but we would need a change to the solr-init. As I'm no Solr expert (and neither a container expert) I'm very glad Rob provided us a full fledged solution on docker hub. 

So, let's try this. If you want to do this as well, read his post to understand the options for the solr-init image. His example is for XM - so we have to make a few small changes to get this working for XMCloud. It will result in a solr-config yml section like this:
solr-init:
    isolation: ${ISOLATION}
    image: rahnemann/solr-init:1.0-ltsc2019
    environment:
      TOPOLOGY: xm-sxa
      SITECORE_SOLR_CONNECTION_STRING: http://solr:8983/solr
      SOLR_CORE_PREFIX_NAME: ${SOLR_CORE_PREFIX_NAME}
      ADDITIONAL_SITECORE_CORES: _horizon_index
    volumes:
      - type: bind
        source: ${LOCAL_DATA_PATH}\solr
        target: c:\solr
So we are using the image from rahnemann here, we kept the topology xm-sxa as that fits our purpose but we have to add an additional Sitecore core for (oh boy) "Horizon". This is the Pages editor (which is not Horizon, but actually is).

You will also need to change the solr mode and the solr connectionstring:
solr:
    ...
    environment:
      SOLR_MODE: standalone
cm:
    ...
    environment:
      ...
      Sitecore_ConnectionStrings_Solr.Search: http://solr:8983/solr
    ...
After applying these changes, I can now run the containers again. More than once. 

It doesn't feel like a very decent solution. And I guess lots of people will not have these issues - but as some people did and blogged about it I would also assume there are more out there who actually do bump into this and maybe this post about my experience can save you some time of you do.