In certain occasions when using the remote provider, such as after performing a system import, the access client may be reloaded, and the remote binary provider's connection pool with the other nodes binarystore will be shut down. However, Artifactory won't be able to recover and the binary provider won't be able to connect to the new access client unless a restart is performed. Artifactory's remote binary store should better handle when the access client is reloaded or recreated.
This issue is not only related to imports, if any loss of database connectivity takes place the issue may occur.
Steps to reproduce:
-setup cluster s3 HA cluster
-perform a system import (with export from another cluster)
-after the import, try to deploy a file. On the node that the file was deployed to, see binary provider errors:
2019-07-29 22:38:00,783 [http-nio-8081-exec-9] [INFO ] (o.a.e.UploadServiceImpl:401) - Deploy to 'example-repo-local:crc-32-1.2.0.tgz' Content-Length: 1782
2019-07-29 22:38:00,820 [http-nio-8081-exec-9] [ERROR] (o.a.a.f.m.ShardingBinaryProviderImpl:326) - Failed to stream binary to sub provider: java.lang.IllegalStateException: Connection pool shut down
java.util.concurrent.ExecutionException: java.lang.IllegalStateException: Connection pool shut down