I don't have a list of affected kernels, particularly since so many different kernel versions were patched. I know the problem was the initial fix for CVE-2017-1000364, so you could check to see if the kernel version has that in it.   I believe many Linux vendors are working on a fix for the fix, e.g.  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=865549 so this advice may become incorrect once the problem is resolved. ... View more

The CDH5.8 version of impala-udf-devel should work with CDH5.9 and CDH5.10, so continuing to use that seems fine. The UDF interface didn't really change in those releases.   The issue was fixed in CDH5.11 so once you get beyond CDH5.10 you can just use the matching versions (as was originally intended). Sorry for the inconvenience.   We did look at backporting the fixes to the 5.9.x and 5.10.x release series to avoid this pain but the fix required a change to our compiler toolchain, which we wouldn't normally do in a maintenance release. ... View more

"-Xss1280k" seems to be sufficient. The default is 1024k I believe, and previously that was always sufficient in our testing.   The crash was caused by a change to the linux kernel that modified the memory layout around thread stacks. As a result with the default Java stack size, the JVM somehow ends up accessing invalid memory. Increasing the stack size mitigates this.t ... View more

Yes, a lot of people have been hitting this after upgrading their kernels! Thank you for following up and confirming that you were able to fix the problem. ... View more

This is https://issues.apache.org/jira/browse/IMPALA-5578   I think you will probably also need to update "Impala Catalog Server Environment Advanced Configuration Snippet (Safety Valve)" before you restart the catalog daemon. ... View more

We'd expect Kudu to be slower than Parquet on a pure read benchmark, but not 10x slower - that may be a configuration problem. We've published results on the Cloudera blog before that demonstrate this: http://blog.cloudera.com/blog/2017/02/performance-comparing-of-different-file-formats-and-storage-engines-in-hadoop-file-system/    Parquet is a read-only storage format while Kudu supports row-level updates so they make different trade-offs. I think we have headroom to significantly improve the performance of both table formats in Impala over time. E.g. in Impala 2.9/CDH5.12 IMPALA-5347 and IMPALA-5304 improve pure Parquet scan performance by 50%+ on some workloads, and I think there are probably similar opportunities for Kudu.   ... View more

I'd suggest looking at the log files for the Impala Daemon role that failed. There will typically be an explanation for the failed startup in there. ... View more