*Closed*: Cloudbreak on Azure: Linux kernel version and impact on kerberizing

[#1]

Hello,

Issue:
Around mid-November 2017, ran into an issue with datanodes not coming up, and namenode therefore not coming out of safe node when we kerberized (MIT kerberos KDC) a cluster after we provisioned via Cloudbreak.

Reference:
https://issues.apache.org/jira/browse/HDFS-12029
https://access.redhat.com/errata/RHBA-2017:1674
https://community.hortonworks.com/articles/109940/after-os-patching-all-the-datanodes-nodes-are-up-b...

Temporary Fix:
Hortonworks support was consulted, and they applied a temporary fix:
Updated hadoop-env.sh configs in HDFS configs in ambari as follows:

Original entry:

export HADOOP_OPTS="-Djava.net.preferIPv4Stack=true
${HADOOP_OPTS}"

With fix:

export HADOOP_OPTS="-Djava.net.preferIPv4Stack=true -Xss1280k ${HADOOP_OPTS}”

Restarted services and we were good to go.

Strategic solution:
Upgrade the underlying OS image to a version that includes a patch.
3.10.0-514.26.2.el7.x86_64 version, per the support engineer

Reporting this since an upgrade is being worked, should have done it earlier 😞

@pdarvasi

[#2 · Accepted solution]

@Airawat

I've just double-checked the new Cloudbreak 1.16.5 version in Marketplace and it seems to have the correct version:

Could You please verify? Thanks!

[#3]

@pdarvasi
Will try out and get back to you. Thanks so much!

[#4]

@Airawat

Have you had the chance to have a look at this one? Thanks!

[#5]

@pdarvasi - testing it today.

[#6]

@pdarvasi:
(1) Availability set (1 for multiple host groups - masters): Works fine
(2) Linux kernel version: is the one in your screenshot - version with patch
(3) Kerberos enabled post-provisioning against MIT Kerberos KDC: No issues
(4) Kerberos enabled at provision time with new MIT Kerberos KDC: Did not complete in 12 hours..lots of alerts, install pending, and services in invalid state and unable to start, Ambari version could not be determined...will open a separate post

[#7]

@Airawat If you consider your original question answered, would you please consider accepting the answer? Thanks!