0
votes

I'm running a 3-node cluster in AWS. Yesterday, I upgraded my cluster from DSE 4.7.3 to 4.8.0.

After the upgrade, the datastax-agent service is no longer registered and the /usr/share/datastax-agent/conf folder has been removed.

PRE-UPGRADE: $ ls -alr total 24836 drwxrwxr-x 3 cassandra cassandra 4096 Aug 10 14:57 tmp drwxrwxr-x 2 cassandra cassandra 4096 Aug 10 14:56 ssl drwxrwxr-x 2 cassandra cassandra 4096 Sep 28 15:14 doc -rw-r--r-- 1 cassandra cassandra 25402305 Jul 14 18:55 datastax-agent-5.2.0-standalone.jar drwxrwxr-x 2 cassandra cassandra 4096 Sep 28 18:23 conf drwxrwxr-x 3 cassandra cassandra 4096 Sep 28 18:13 bin drwxr-xr-x 118 root root 4096 Oct 2 18:02 .. drwxrwxr-x 7 cassandra cassandra 4096 Oct 7 19:03 .

POST-UPGRADE: $ ls -al total 24976 drwxr-xr-x 3 cassandra cassandra 4096 Oct 5 20:45 . drwxr-xr-x 114 root root 4096 Oct 5 18:23 .. drwxr-xr-x 3 cassandra cassandra 4096 Oct 5 20:45 bin -rw-r--r-- 1 cassandra cassandra 25562841 Sep 10 20:43 datastax-agent-5.2.1-standalone.jar

Also, /etc/init.d/datastax-agent file has been deleted. I don't know how I'm supposed to start/stop the service now.

Can I restore the files from the rollback directory? What effect will that have?

1
What method of installation where you using? Packages, tar ball, standalone installer? What OS version is this? - Sven Delmas
standalone installer DataStaxEnterprise-4.8.0.2015091812-linux-x64-installer.run $ lsb_release -a No LSB modules are available. Distributor ID: Ubuntu Description: Ubuntu 14.04.3 LTS Release: 14.04 Codename: trusty - LHWizard
I'm not sure why the .run installer would blow away your agent install. Have you tried clicking the fix button in OpsCenter to re-install the agent? Otherwise you can use the package manager to install it. - phact
On further digging, it appears that the 4.8 installer moved the whole folder structure from /usr/share/datastax-agent to /usr/share/dse/datastax-agent/ - LHWizard
That folder under dse is just there as a fallback. The new installer is using the package install internally (using it's directory structure) to allow opscenter to properly manage the agent. It seems that that install failed... What happens when you try a dpkg install manually with the included package? Do you have the log file? Can you share that with me? - Sven Delmas

1 Answers

2
votes

In this particular case what happened was that the dpkg install found a preexisting /etc/init.d/datastax-agent file and only put /etc/init.d/datastax-agent.fpk.bak into place. A "sudo dpkg -P datastax-agent" followed by a "sudo dpkg -i /usr/share/dse/datastax-agent/datastax-agent_5.2.1_all.deb" fixed the issue. We had to first kill the already running agent processes and then do a service restart.

Will investigate how that could have happened... that's still a little bit of a mystery to me.