You have to increase the spark.network.timeout
value too.
The documentation clearly states:
spark.executor.heartbeatInterval
should be significantly less than
spark.network.timeout
It's hard to interpret "significantly less" but by default spark.network.timeout
is 12X greater (120s) than spark.executor.heartbeatInterval
(10s).
Try:
--conf spark.network.timeout 100000 --conf spark.executor.heartbeatInterval 10000
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…