CODEWITHSUNDEEP
cassandradatastaxdatastax-enterpriseopscenter
pdeva
pdeva

Reputation: 45491

upgrading cassandra with DataStax Lifecycle manager

The DataStax Opscenter LifeCycle Manager only seems to have an option to run an 'install' job. Looking at the language, it seems to be only to provision new nodes.

Can LifeCycle Manager be used to upgrade existing (managed) clusters to newer version of Datastax enterprise?

Upvotes: 0

Views: 292

Answers (1)

Mike Lococo
Mike Lococo

Reputation: 684

Edit 2018-05 OpsCenter 6.5.0 has been released and provides assistance in the process of upgrading DSE between patch releases... aka going from DSE 5.0.3 to 5.0.6. Docs and https://docs.datastax.com/en/opscenter/6.5/opsc/LCM/opscLCMjobsOverview.html and https://docs.datastax.com/en/opscenter/6.5/opsc/LCM/upgradeDSEjob.html.

DataStax engineer here, I work on Lifecycle Manager. Currently LCM cannot help you upgrade nodes, and while I'm not able to share information about future roadmap and unreleased features, I can say we know that customers want to use LCM for upgrades and we agree that it would be a valuable feature.

As of OpsCenter 6.1.x, you must manually upgrade your nodes, and then update your LCM configs to match the new versions. From that point onward you can use LCM for install/config jobs in the upgraded cluster. This isn't a detailed howto, but broadly:

  1. Review the upgrade guide so you know what needs to be done: https://docs.datastax.com/en/upgrade/doc/upgrade/datastax_enterprise/upgrdDSE.html
  2. Perform the upgrade manually, outside of LCM. Note that if you use apt to manage packages, and are not upgrading the very most recent version available, you'll have to use a pretty giant apt-command in order to work around a dependency resolution in apt when upgrading to an "old" version. The resulting command will look something like: apt-get install -y -qq -o Dpkg::Options::=--force-confdef -o Dpkg::Options::=--force-confold dse-pig=5.0.11-1 dse-libhadoop2-client=5.0.11-1 dse-libspark=5.0.11-1 dse-libhadoop-native=5.0.11-1 dse-libmahout=5.0.11-1 dse-hive=5.0.11-1 dse-libpig=5.0.11-1 dse-libsolr=5.0.11-1 dse-libgraph=5.0.11-1 dse-libtomcat=5.0.11-1 dse-libhadoop=5.0.11-1 dse-libhive=5.0.11-1 dse-full=5.0.11-1 dse-libcassandra=5.0.11-1 dse=5.0.11-1 dse-libsqoop=5.0.11-1 dse-libhadoop2-client-native=5.0.11-1 dse-liblog4j=5.0.11-1
  3. Once the manual upgrade is complete, you'll temporarily be in a position where LCM jobs cannot run successfully, since the version of DSE installed does not match the version of DSE that LCM is configured to deploy. At this point LCM jobs will fail with a DSE version mismatch error. To fix this, proceed...
  4. Clone your configuration profile (which is associated with your old dse version) to a new CP using the new DSE version. If you're doing a patch upgrade, this will be pretty simple. If you're doing a major upgrad via the API, you need to be very careful to remove config parameters that DSE no longer supports.
  5. Edit your cluster model so that the cluster, plus any datacenters or nodes with CP's defined use your newly cloned CP for your current datastax version instead of your old cp for your old datastax version. At this point, you can brought LCM back into sync with your cluster. You can proceed to run install/configure jobs again.

Not a simple procedure, but it is possible to upgrade your cluster outside of LCM and then sync lcm up with the new config so you can continue managing it from there. As previously noted, we understand this is not a simple process and understand that there's significant value in providing LCM upgrades natively.

Upvotes: 2

Related Questions