[ https://issues.apache.org/jira/browse/BEAM-991?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15711705#comment-15711705 ]
Joshua Fox commented on BEAM-991: --------------------------------- Daniel Halperin, unfortunately I cannot participate too much in the development and full testing, both for the usual reason of time commitments but also because I am not close enough to the infrastructure to fully understand what is happening, particularly in edge cases. However, I will be glad to serve as an early user. My situation: I wrote a Datastore backup in Dataflow, after Google's Backup and Managed Backup tool failed on various bugs. However, this new tool cannot copy any Kind with Item size >20 KB (max in Datastore is 1 MB). So, I can only use it to backup Kinds with average Item size 10 KB (since there may be variation.) I wrote a second a simple multithreaded but nondistributed backup tool, using "select-insert" loops. It works, but of course is less scalable and more expensive than the Dataflow tool. I use a combination of the Dataflow and the nondistributed tool now, for different Kinds, but can easily switch to using just Dataflow. > DatastoreIO Write should flush early for large batches > ------------------------------------------------------ > > Key: BEAM-991 > URL: https://issues.apache.org/jira/browse/BEAM-991 > Project: Beam > Issue Type: Bug > Components: sdk-java-gcp > Reporter: Vikas Kedigehalli > Assignee: Vikas Kedigehalli > > If entities are large (avg size > 20KB) then the a single batched write (500 > entities) would exceed the Datastore size limit of a single request (10MB) > from https://cloud.google.com/datastore/docs/concepts/limits. > First reported in: > http://stackoverflow.com/questions/40156400/why-does-dataflow-erratically-fail-in-datastore-access -- This message was sent by Atlassian JIRA (v6.3.4#6332)