Hi, Joe, you're right, these files do not exist. These are special "trash" files. In IGFS deletion is performed in 2 phases: 1st a file is moved to special "trash" folder (which does not have normal path and thus cannot be found on the file system), and 2nd -- deletion of the trash entries performed in the background (IgfsDeleteWorker does that, see org.apache.ignite.internal.processors.igfs.IgfsDeleteWorker#delete(org.apache.ignite.lang.IgniteUuid, java.lang.String, org.apache.ignite.lang.IgniteUuid)). The name of the file you see in the log (containing "|" character) composed with operation org.apache.ignite.internal.processors.igfs.IgfsUtils#composeNameForTrash . I don't see obvious reason why these transactions may fall into the loop after a node rejoin. I think , we should create a ticket for that problem . What can help us to track down and fix the issue is a minimal possible scenario that reproduces the problem.
-- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/igfs-meta-behavior-when-node-restarts-tp13155p13283.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.
