Help with understanding Solr Master Slave sync

[Former Member]

I have a situation where I am working with Hybris Commerce setup in AWS.

I have 2 BackOffice servers. 2 FrontOffice Servers 1 Solr Master 2 Solr Slaves.

One of the Slaves is never correctly indexed. So on the site, one test shows facets and then again randomly not.

I temporarily fixed this by removing the faulty slave.

Then it worked. I recreated the slave, ran a full index, and again it is not in sync.

What is a good resource to gain an in depth understanding of Hybris with Solr. For example here is some assumptions in my present understanding, as I am battling to solve this issue I think my assumptions may be false. 1) I assume the master is not directly read by any FrontOffice servers. 2) When a full index cron job is run I assume that Hybris has PUSHED the data to the Master only. So the Solr Master does not pull data from Hybris. Or does it. 3) I am assuming the Slaves PULL information from the master. 4) The Master never PUSHES data to the slaves. 5) I assume their should be one Slave for each FrontOffice server. 6) From above, if one Slave is in sync with Master and the other is not, then there could be a network issue between them.

Regards Capetown South Africa

[arvind-kumar_avinash · Accepted solution]

Hi - please find below the answers:

• 1) I assume the master is not directly read by any FrontOffice servers. - [Y]

• 2) When a full index cron job is run I assume that Hybris has PUSHED the data to the Master only. So the Solr Master does not pull data from Hybris. - [Y]

• 3) I am assuming the Slaves PULL information from the master. - [Y]

• 4) The Master never PUSHES data to the slaves. - [Y]

• 5) I assume their should be one Slave for each FrontOffice server. - [Not necessary e.g. if you may have 10 FrontOffice nodes and only 2 Solr Slave nodes.]

• 6) From above, if one Slave is in sync with Master and the other is not, then there could be a network issue between them. - [I can not say anything for sure without proper investigation. Did you try to restart (stop and start) the faulty slave server? In addition to this, you can configure one more slave and check whether the new slave is also out of sync.]

Here some reference documents that may be useful to you:

1. https://wiki.hybris.com/display/hybrisALF/Solr+Infrastructure

2. https://wiki.hybris.com/display/hybrisALF/Solr+Troubleshooting

3. https://help.hybris.com/1808/hcd/f64b31a2c8274d3eae8a53abf206f247.html

Best regards.

[Former Member]

Thank you Geff, and Arvind. Both your answers have helped me greatly.

[arvind-kumar_avinash]

You are most welcome.

[geffchang]

Check the official Solr documentation on Index Replication: https://lucene.apache.org/solr/guide/6_6/index-replication.html

The Slaves simply poll (and pull) from Master. The configuration for Slave is different from master, but the configuration for multiple Slaves can be the same, if not similar.

Aside from configuring Solr, you also need to make sure that you have configured Solr on Hybris. Make sure that all the SolrEndpointUrl instances are correct.