Details
Major Description
In some recent upgrade testing, we saw this error causing the NodeManager to fail to startup afterwards:
org.apache.hadoop.service.ServiceStateException: com.google.protobuf.InvalidProtocolBufferException: Protocol message contained an invalid tag (zero). at org.apache.hadoop.service.ServiceStateException.convert(ServiceStateException.java:105) at org.apache.hadoop.service.AbstractService.init(AbstractService.java:173) at org.apache.hadoop.service.CompositeService.serviceInit(CompositeService.java:108) at org.apache.hadoop.yarn.server.nodemanager.NodeManager.serviceInit(NodeManager.java:441) at org.apache.hadoop.service.AbstractService.init(AbstractService.java:164) at org.apache.hadoop.yarn.server.nodemanager.NodeManager.initAndStartNodeManager(NodeManager.java:834) at org.apache.hadoop.yarn.server.nodemanager.NodeManager.main(NodeManager.java:895) Caused by: com.google.protobuf.InvalidProtocolBufferException: Protocol message contained an invalid tag (zero). at com.google.protobuf.InvalidProtocolBufferException.invalidTag(InvalidProtocolBufferException.java:89) at com.google.protobuf.CodedInputStream.readTag(CodedInputStream.java:108) at org.apache.hadoop.yarn.proto.YarnSecurityTokenProtos$ContainerTokenIdentifierProto.<init>(YarnSecurityTokenProtos.java:1860) at org.apache.hadoop.yarn.proto.YarnSecurityTokenProtos$ContainerTokenIdentifierProto.<init>(YarnSecurityTokenProtos.java:1824) at org.apache.hadoop.yarn.proto.YarnSecurityTokenProtos$ContainerTokenIdentifierProto$1.parsePartialFrom(YarnSecurityTokenProtos.java:2016) at org.apache.hadoop.yarn.proto.YarnSecurityTokenProtos$ContainerTokenIdentifierProto$1.parsePartialFrom(YarnSecurityTokenProtos.java:2011) at com.google.protobuf.AbstractParser.parsePartialFrom(AbstractParser.java:200) at com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:217) at com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:223) at com.google.protobuf.AbstractParser.parseFrom(AbstractParser.java:49) at org.apache.hadoop.yarn.proto.YarnSecurityTokenProtos$ContainerTokenIdentifierProto.parseFrom(YarnSecurityTokenProtos.java:2686) at org.apache.hadoop.yarn.security.ContainerTokenIdentifier.readFields(ContainerTokenIdentifier.java:254) at org.apache.hadoop.security.token.Token.decodeIdentifier(Token.java:177) at org.apache.hadoop.yarn.server.utils.BuilderUtils.newContainerTokenIdentifier(BuilderUtils.java:322) at org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl.recoverContainer(ContainerManagerImpl.java:455) at org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl.recover(ContainerManagerImpl.java:373) at org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl.serviceInit(ContainerManagerImpl.java:316) at org.apache.hadoop.service.AbstractService.init(AbstractService.java:164) ... 5 more
The NodeManager fails because it's trying to read a ContainerTokenIdentifier in the "old" format before we changed them to protobufs (YARN-668). This is very similar to YARN-5594 where we ran into a similar problem with the ResourceManager and RM Delegation Tokens.
To provide a better experience, we should make the code able to read the old format if it's unable to read it using the new format. We didn't run into any errors with the other two types of tokens that YARN-668 incompatibly changed (NMTokenIdentifier and AMRMTokenIdentifier), but we may as well fix those while we're at it.
