The web interface that allow users to consult and visualize data collected for their executed jobs on Marconi cluster can be reached at the following address:
This is based on Grafana, an open source analytics & monitoring solution (www.grafana.com).
All users with active projects on Marconi-A3 can request the access to the web interface by writing an e-mail to superc@cineca.it.
The user will be then associated to FUSIO_hpcmd_ud project, and once this association will be effective, the user will be able to login to the site by following the "Sign in with CINECA-HPC" button and using her/his HPC credentials (the same username and password used to login to HPC clusters):
Please be aware that 2FA is enabled and, if not done yet, you will need to activate it and configure the OTP:
This is the User's Space main page.
Contains:
- information about the content of the site; as this site is under construction, metrics, dashboards, panels or functionalities might be included or removed in the future. Updates will be reported in this page.
- links to Dashboards where users can explore data
- Information about available command line tools
- Other information for executed applications
- To explore the activity on the cluster:
- select a time interval to visualize data:
- and expand the first row in the dashboard:
- To visualize raw data for your jobs:
- Select a time interval to generate the list of jobs that have their start time in this period
- Select a jobid from the list
- Select the metric to visualize: perf, gpfs, exe, memory, network
- Expand the second row in the dashboard to visualize generated panels
- To get additional information about the specific command that generated raw data consult the command info dashboard and expand the row of interest to visualize the content:
- Select a time period to visualize your jobs that have their end time in the reference period
- Jobs that used from 1 to 4 nodes - follow the links provided in the table to open specific information dashboard:
- executable info: stats provided by the ps command line tool, the /proc/loadavg file, the /proc/{pid}/maps file
- perf dashboard: events information collected by the Linux perf tool (collected data grouped by socket and by host)
- gpfs dashboard: GPFS filesystem info collected by the mmpmon tool
- network dashboard: network info collected by the opainfo tool.
- Jobs that used more than 4 nodes - follow the links provided in the table to open specific information dashboard:
- executable info: stats provided by the ps command line tool, the /proc/loadavg file, the /proc/{pid}/maps file
- perf dashboard: events information collected by the Linux perf tool (collected data grouped by socket)
- gpfs dashboard: GPFS filesystem info collected by the mmpmon tool
- network dashboard: network info collected by the opainfo tool.
- Dashboards specific for metrics:
- Select a time period to visualize your jobs that have their end time in the reference period to get:
- the node-h or core-h consumed by the jobs
- the total number of executed app in the reference period
This service is in pre-production phase.
Please write an e-mail to superc@cineca.it, by including the subject tag [hpcmd], to report any issues you experienced with the web interface or your comments or suggestions that will help us to improve this service.