Enabling debug logging level in class org.apache.ignite.internal.processors.igfs.IgfsImpl can give tracing of all the IGFS operations performed -- this can help to track this issue down. Many operations in IGFS are implemented using "retry" logic , like the following
while (true) { try (GridNearTxLocal tx = startTx()) { // Lock participants. Map<IgniteUuid, IgfsEntryInfo> lockInfos = lockIds(allIds); if (!checkSomeIntegrity(....)) continue; // do smth .... tx.commit(); } } , so they may experience contention (staring and abandoning a transaction many times per one operation) if some overlapping operations are attempted concurrently. -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/igfs-meta-behavior-when-node-restarts-tp13155p13206.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.