Probably one of the protobuf incompatibility. Unfortunately we don't have an open source tool to detect protobuf incompat.
A few related issues: 1. HDFS-15700 <https://issues.apache.org/jira/browse/HDFS-15700> 2. 1. HDFS-14726 <https://issues.apache.org/jira/browse/HDFS-14726> 2. 1. HDFS-13371 <https://issues.apache.org/jira/browse/HDFS-13371> 1. HDFS-15660 <https://issues.apache.org/jira/browse/HDFS-15660> 2. I know folks upgraded from 2.7 to 2.10 (LinkedIn?), and 2.8 to 2.10 (Verizon Media). 3. Searched JIRA I don't see a known bug between 2.6 and 2.10 in the DN heartbeat protobuf. On Mon, Dec 14, 2020 at 10:57 PM Chad William Seys <[email protected]> wrote: > Hi all, > Is it required or highly recommended that one not skip between HDFS > (hadoop) versions? > I tried skipping from 2.6 to 2.10 and it didn't work so well. :/ > Actually, I tested this with a tiny cluster and it worked, but on the > production cluster the datanodes did not report blocks to the namenodes. > (Did report storage and connectivity otherwise.) > > Chad. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
