Due to the large amount of annual index data, we plan to divide 20 shards for clustering
Hardware: (32G memory, 8-core CPU, each server mounts 4 500G ssd hard disks) X 5 sets
There are currently two ways to build:
1. Each solr instance corresponds to a disk, and a total of 20 instances need to be opened
solr -c -m 8g -z xx1:2181,xx2:2181,xx3:2181,xx4:2181,xx5:2181 -s /mnt/disk1/data/solrcloud/solr1 -p 8941
solr -c -m 8g -z xx1:2181,xx2:2181,xx3:2181,xx4:2181,xx5:2181 -s /mnt/disk2/data/solrcloud/solr2 -p 8942
...
After 20 instances are created, use the following solr command:
http://xxx:xxx/solr/admin/collections?action=CREATE&name=uuwatch-year2016&numShards=20&router.field=group
2. If you only want to open 5 instances, you can transform it on the basis of the above (tested, this method is more efficient for indexing and retrieval)
1). Establish a soft link in the /home/wp/data directory, and link to the index folder under solr.home corresponding to the 20 previously created instances.
The directory structure of the new solr.home is as follows:
2) Do the above operations in sequence on other machines
3) Stop the service of all machines
4) Execute the following command on each machine
solr -c -m 28g -z xx1:2181,xx2:2181,xx3:2181,xx4:2181,xx5:2181 -s /home/wp/data -p 8941
Finally, as follows (tested, document routing works fine too):