Solving Common SSL-Related Issues with IBM Spectrum Conductor Deep Learning Impact 1.1
sco1 2700065TNJ Visits (7074)
Authors: Sum Huynh, Faye Fang
Published on February 22, 2018 / Updated on February 23, 2018
By default, the cluster management console in IBM Spectrum Conductor Deep Learning Impact is SSL-enabled. It is unlikely for SSL related issues to occur when IBM Spectrum Conductor Deep Learning Impact 1.1 is first installed. However, over time, changes to your system can stop your browser from accessing IBM Spectrum Conductor Deep Learning Impact services due to SSL-related issues shown in screen captures below. These issues can appear after you install a new browser on a desktop, clear existing browser certifications, or change network related configurations on the management nodes. This blog walks you through a scenario where you encounter such issues and shows you several steps that you can take to help you troubleshoot and resolve them. In this blog, Firefox 58.0.2 (64-bit) is used.
First, make sure that you can access the cluster management console using a fully qualified domain name (FQDN) and default port, such as: http
Or, if you click the Workloads tab, and navigate to Spark > Deep Learning, you might see the following SSL-related error:
The following steps will take you through a series of steps to troubleshoot these types of issues.
As suggested by the error messages in the screen captures above, make sure that the EGO services are started. To check EGO services, log into the cluster management console, and navigate to System & Services > EGO Services > Services as shown in the next screen capture:
If ascd and dlpd services had been started as shown in the screenshot then this is not the cause of the problem. If these services are not in started state, click on the pull-down menu ‘Actions’ on the service to start the service.
After verifying that services are started, make sure that you can reach and authenticate to the services. Here, you can access the REST services directly and enter your IBM Spectrum Conductor with Spark credentials as needed. The screenshot below shows accessing IBM Spectrum Conductor Deep Learning Impact directly on default port 9243. You can also repeat this step for IBM Spectrum Conductor with Spark REST service. For example, using default port that would be http
If you can reach the REST services, as shown below, then this is not the cause of the problem.Now, go back to the cluster management console and refresh the tab. Refreshing the tab, you might see a warning that the connection is not secure:
If you click “Advanced” then “Add Exception …”, then you might see this:
In addition to the warning messages, the URL in this screen shows a short name where FQDN is expected – this is a hint to the problem.
To make sure that the cluster is configured with FQDN, log into the management node and check using egosh command as shown below. As shown in the following screenshot, the master host name is FQDN so this is not the cause of the problem.
Next, you can run the egosh client command to check whether FQDN or short name is used:
As shown in the screen capture, the host name in the URL is short name where FQDN is expected. To confirm, you can run the hostname command as follows:
Indeed, the command ‘hostname –f’ should have shown FQDN, not short name. This is the cause of the problem.
You should then check the /ect/hosts file on the management node. It is possible that you might have to edit this file and for example reverse the order of the short name and FQDN for the entry for the current host.
Once the command hostname –f returns FQDN, then you can proceed to verify the egosh client view command again as follows:
# egosh service stop all
# egosh service start all
# egosh client view REST_BASE_URL_1
Hopefully when you refresh your browser, the SSL-related issue is no longer there:
SSL-related issues can take time to troubleshoot as there are several places you will need to try - while paying detailed attention to things such as FQDN and short name. Hopefully the steps in this blog have helped you troubleshoot your SSL-related issues. If you have any questions regarding this or other SSL issues, feel free to join our Slack channel and ask us directly. Join
For more information about IBM Spectrum Conductor Deep Learning Impact, see: