Restriction

Last Updated: Jun 21, 2016

  • The number of resources quoted by a single job can not exceed 256. Table and archive are considered as one unit.
  • The size of total bytes quoted by a single job can not exceed 512 M.
  • The number of input channels of a single job cannot exceed 1024 (the number of input tables cannot exceed 64.); the number of output channels of a single job cannot exceed 256.
  • The label specified in multiple outputs cannot be null or a empty string. Its length cannot exceed 256 and it can only contain A-Z, a-z, 0-9, _, #, - and so on.
  • The number of counters defined in a single job cannot exceed 64. The group name and counter name cannot contain ‘#’ and the length sum of them cannot exceed 100.
  • The worker number of a single job is calculated by framework and the maximum is 1000, otherwise the expection will be thrown.
  • The default CPU size occupied by a single job is 200 and the range is [50, 800].
  • The default memory size occupied by a single job is 4096 and the range is [256M, 12G].
  • The occurrence that a single worker reads a resource repeatedly cannot be more than 64.
  • The default plit size is 64M, which can be set by users and the range is: 0 < split_size <= (9223372036854775807 >> 20);
  • In the cluster running process, GraphLoader/Vertex/Aggregator in MaxCompute Graph program are limited by Java sandbox. (the main program of Graph job has no this limitation.) For details, refer to Java Sandbox.
Thank you! We've received your feedback.