Thank you for all of the suggestions > - A couple places seem to keep a Cloudera copyright/license - eg > common/thrift/beeswax.thrift, FrontEndTestBase.java, etc. (I grepped for > Cloudera and found these). Should probably check over these and fix before > the "real" RC.
Reopened https://issues.cloudera.org/browse/IMPALA-3918 > - I tried building using 'buildall.sh' but it failed with 'fatal: Not a git > repository' when trying to do a git clean. Maybe some tweaks need to be > made so that Impala can be built from a tarball? I worked around it using > 'git init' inside my extracted directory. https://gerrit.cloudera.org/#/c/4336/ When this is in master, I'll cherry-pick it to branch-2.7.0 > - would be nice if the version number didn't have 'cdh5' in it (eg > impala-shell-2.7.0-cdh5-INTERNAL, seems to come from bin/version.info, > bin/save-version.sh, etc). Should probably be '2.7.0-incubating' Lars is working on this now: https://gerrit.cloudera.org/#/c/4187/3 Some possibilities for 2.7.0: 1. Leave it. Pros: expedient; Cons: inaccurate, not Apache 2. Wait until we can change this is master without confusing existing tools. Pros: clean; Cons: delay in releasing 3. Make a commit just to branch-2.7.0. Pros: Doesn't confuse existing tools; Cons: Makes branch-2.7.0 not a pure subset of master. What does everyone think? > - can't seem to use testdata/bin/run-all.sh to start DFS. I get: > Starting hdfs (Web UI - http://localhost:5070) > Failed to start hdfs-datanode. The end of the log > (/tmp/incubator-impala/testdata/cluster/cdh5/node-2/var/log/hdfs-datanode.out) > is: > Failed to start hdfs-datanode. The end of the log > (/tmp/incubator-impala/testdata/cluster/cdh5/node-3/var/log/hdfs-datanode.out) > is: > Failed to start hdfs-datanode. The end of the log > (/tmp/incubator-impala/testdata/cluster/cdh5/node-1/var/log/hdfs-datanode.out) > is: > > That said, it appears that impalad built OK (with build-all.sh -notests) I could fix this by making sure IMPALA_HOME was set correctly and sourcing bin/impala-config.sh. > some nits/suggestions (ok to address in a later release: > - a few mentions of 'Cloudera Impala' in the various pom files Opened https://issues.cloudera.org/browse/IMPALA-4093 > - would be great if buildall.sh could check if there is enough remaining > space on the drive before building. I had 20GB free but still ran out of > space trying to do the default build (had to rebuild with -notests) Opened https://issues.cloudera.org/browse/IMPALA-4094 > - consider setting up a cloudfront distribution in front of the > native-toolchain S3 bucket? It downloads pretty slowly for me Filed https://issues.cloudera.org/browse/IMPALA-4095 > -- related: consider stripping debug info from some of the built deps? eg > Kudu is 500+MB which seems unnecessary for a test dependency. Which object is this? I don't find it. $ find . -iname '*kudu*' -size +100M | wc -l 0 Thanks again, Todd! Your input as our mentor has been invaluable.