Add a section in the diskbalancer documentation on how to report issues and how to debug diskbalancer usage.
SUCCESS: Integrated in Jenkins build Hadoop-trunk-Commit #10437 (See https://builds.apache.org/job/Hadoop-trunk-Commit/10437/)
HDFS-10562. DiskBalancer: update documentation on how to report issues (aengineer: rev 1c0d18f32289837b8981aed80e7bdcd360adfb85)
Arpit Agarwal Thanks for the code reviews and your comments. I have committed this to trunk.
+1 thanks Anu Engineer.
This message was automatically generated.
Fixed the issue pointed out by Arpit Agarwal
Thanks for catching that, will do.
+1. A minor comment:
212 *max disk errors* - The configuration used for this move step. currently it will
213 report it as zero, since the user interface to control these values per step
214 is not in place. It is a future work item. The default or the command line
215 value specified in plan command is used for this value.
The sample output shows "maxDiskErrors" : 5,. Should we fix the output to show zero to be consistent with the description?
Arpit Agarwal Thanks for the code review comments. I have updated the patch.
Hi Anu Engineer, thanks for the doc improvements. One question regarding the bandwidth. Should the first sentence be reworded to say something like This is the maximum aggregate source disk bandwidth used by the balancer?
223 *bandwidth* - From disk balancer's point of view this is the bandwidth of the
224 source disk. After moving a block disk balancer computes how many seconds it
225 should have taken to move that block with the specified bandwidth. If the
226 actual move took less time than expected, then disk balancer will sleep for
227 that duration. Please note that currently all moves are executed
228 sequentially by a single thread.