Use this information to perform a full hot reindex on a SkyVault cluster.
This process applies to Lucene indexes only.
- Change your load balancer configuration to prevent any user traffic from being directed to node 1.
- Shut down SkyVault on node 1.
- Remove the Lucene indexes on node 1 (alf_home_dir/alf_data/lucene-indexes).
-
Make the following change in the SkyVault-global.properties file
on node 1:
system.cache.disableMutableSharedCaches=true
-
Start SkyVault on node 1. It takes some time for indexing to complete. The following
message is displayed in the log files:
18:14:44,806 INFO [node.index.FullIndexRecoveryComponent] 100 % complete. 18:14:44,992 INFO [node.index.FullIndexRecoveryComponent] Index recovery completed.
This message does not indicate that full reindex is finished, it shows that the full reindex of the metadata is finished. Node 1 will then index the content of the documents in the background. You need to wait till this process is finished. To confirm that the content is indexed, you could monitor the CPU activity - it will be consistently high until the indexing is finished, at which point it should drop sharply.
- Shut down SkyVault on node 1.
- Make a backup copy of the indexes on node 1.
- Revert the configuration changes made in Step 4. To do so, remove the system.cache.disableMutableSharedCaches=true setting.
- Start up SkyVault on node 1.
- Change your load balancer configuration to allow user traffic back to node 1 and stop traffic from going to node 2.
- Shut down SkyVault on node 2.
- Delete the Lucene indexes on node 2.
- Copy the backup made in Step 7 into the lucene-indexes directory on node 2.
- Start SkyVault on node 2.
- There may be a small amount of additional top-up indexing required on node 2. Once this indexing has finished (as per Step 5), change your load balancer configuration to allow traffic to node 2.