I am simply trying to load an RDD from disk with transposeRowsRDD.avro(baseInputFname).rdd.map( <some pointwise function here, not important>) and I get this error in my log:
16/02/04 11:44:07 ERROR TaskSchedulerImpl: Lost executor 7 on nid00788: Remote RPC client disassociated. Likely due to containers exceeding thresholds, or network issues. Check driver logs for WARN messages. When I check the log for that node (I guess this is what it means by driver?) I see: 16/02/04 11:43:55 INFO TorrentBroadcast: Started reading broadcast variable 152 16/02/04 11:43:55 INFO MemoryStore: Block broadcast_152_piece0 stored as bytes in memory (estimated size 19.3 KB, free 8.8 GB) 16/02/04 11:43:55 INFO TorrentBroadcast: Reading broadcast variable 152 took 4 ms 16/02/04 11:43:55 INFO MemoryStore: Block broadcast_152 stored as values in memory (estimated size 364.3 KB, free 8.8 GB) 16/02/04 11:43:56 INFO MemoryStore: Block rdd_1634_1637 stored as values in memory (estimated size 24.0 B, free 8.8 GB) 16/02/04 11:43:56 INFO Executor: Finished task 1637.0 in stage 0.0 (TID 1637). 2602 bytes result sent to driver 16/02/04 11:43:56 INFO CoarseGrainedExecutorBackend: Got assigned task 1643 16/02/04 11:43:56 INFO Executor: Running task 1643.0 in stage 0.0 (TID 1643) 16/02/04 11:43:56 INFO CacheManager: Partition rdd_1634_1643 not found, computing it 16/02/04 11:43:56 INFO HadoopRDD: Input split: file:/global/cscratch1/sd/gittens/CFSRA/CFSRAparquetTranspose/CFSRAparquetTranspose0/part-r-00007-ddeb3951-d7da-4926-a16c-d54d71850131.avro:134217728+33554432 So the executor seems to have crashed without any error message being emitted, with plenty of memory on hand (I also grepped for WARNING messages, didn't see any). Any idea on what might be happening, or how to debug? Several other executors are also lost with the same behavior. I'm using Spark in standalone mode. -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/cause-of-RPC-error-tp26151.html Sent from the Apache Spark User List mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org