Hello. We have a test ignite 2.5 installation with 1 server and 1 client. Server starts from systemctl service with parameters:
Ignite configuration: Client consistently loading data with IgniteDataStreamer. After those actions: 1. start server 2. start client 3. 3-4 times restart client without proper application shutdown 4. stop client 5. stoped service with systemctl stop After that serever node refused to start with java.nio.BufferUnderflowException: After few investigations in source code I came to conclusion that some of checkpoint files in db/node-XXX/cp were corrupted. Because if BufferUnderflowException is a RuntimeException and doesnt extends IOException, it is not being catched in readPointer method and not being wrapped into IgniteCheckedException. After removing whole cp directory node started without exception. Is it proper solution? Will node startup correctly and it's data become in proper state in that case if it would be a several servers cluster? I tried to google for "Ignite BufferUnderflowException" or "Failed to read checkpoint pointer from marker file" - exception message thrown on IOException when reading cp file, but didn't find any mentioning of that error except source code. -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
