Details
-
Bug
-
Status: Resolved
-
Critical
-
Resolution: Invalid
-
Impala 2.3.0
-
None
-
None
Description
There are several problems with the following job (and the secure one):
http://sandbox.jenkins.cloudera.com/view/Impala/view/Cluster%20Runs/view/10-node-cdh5/job/compute-stats-10node-cdh5/
The job relies on checking out and building the Impala repo. Not only is this unnecessary, it also makes the job slow (only 40% of the time spent actually computing stats) ad brittle. For example, I just had to debug a problem caused by an update to the Impala toolchain. The change in the toolchain seems completely unrelated to the goal of this job. Our perf runs blindly run the compute stats job before any queries, so breaking this job means our perf runs fail.
We should clean this up to not rely on the Impala repo and make this job more reliable.