Data Extraction

29 views
Skip to first unread message

Yuna Joung

unread,
Jul 22, 2016, 4:52:01 PM7/22/16
to cbtool-users
I ran a ycsb workload on EC2 and all seemed to be successful. However, after I detached everything and ran monextract all, the data appeared to be incomplete/empty. What is going on here?
HOST_runtime_os_EXP-07-22-2016-08-30-31-PM-UTC.csv
trace_EXP-07-22-2016-08-30-31-PM-UTC.csv
VM_management_EXP-07-22-2016-08-30-31-PM-UTC.csv
VM_runtime_app_EXP-07-22-2016-08-30-31-PM-UTC.csv
VM_runtime_os_EXP-07-22-2016-08-30-31-PM-UTC.csv

Michael R. Hines

unread,
Jul 22, 2016, 5:08:49 PM7/22/16
to Yuna Joung, cbtool-users
Is this with or without SPEC?

1. If it's with SPEC: You probably have the flag "collect_from_guest"
set to 'false' in the SPEC yaml. Go ahead and toggle it.
2. If it's cbtool by itself, it's still the same variable, but it's
configured like this:

# Drop this into your configuration file:
cbtool/configs/USERNAME_definitions.txt (do not modify the templates
directly).
[MON_DEFAULTS]
COLLECT_FROM_GUEST = $True


HOST metrics are not supported by EC2 anyway, but OpenStack does.

If you have a private cloud, we could guide you on collecting host
metrics if you really need them.

/*
* Michael R. Hines
* Platform Engineer, DigitalOcean.
*/

Yuna Joung

unread,
Jul 22, 2016, 5:27:14 PM7/22/16
to cbtool-users, jo...@college.harvard.edu
This is just cbtool by itself. I put your line in my config file but I am still not getting any data. However, I am seeing this error message:

status: No samples of runtime_app metrics for all VMs were found (the file /home/ubuntu/lampson/lib/auxiliary//../../data/EXP-07-22-2016-09-13-23-PM-UTC/VM_runtime_app_EXP-07-22-2016-09-13-23-PM-UTC.csv will be empty). 
status: No samples of runtime_os metrics for all VMs were found (the file /home/ubuntu/lampson/lib/auxiliary//../../data/EXP-07-22-2016-09-13-23-PM-UTC/VM_runtime_os_EXP-07-22-2016-09-13-23-PM-UTC.csv will be empty). 
status: No samples of runtime_os metrics for all HOSTs were found (the file /home/ubuntu/lampson/lib/auxiliary//../../data/EXP-07-22-2016-09-13-23-PM-UTC/HOST_runtime_os_EXP-07-22-2016-09-13-23-PM-UTC.csv will be empty).

How should I proceed from here?

Michael R. Hines

unread,
Jul 22, 2016, 5:47:17 PM7/22/16
to Yuna Joung, cbtool-users
When something like this happens, it's usually because you have a complex network environment or your security groups are not setup correctly.

The tool monitors everything hierarchically: Each VM reports data to the "lead" VM in the cluster (what we call a load manager VM), and then this lead VM writes that data backwards to mongodb via a direct connection to the host running the tool itself.

If connectivity to the mongodb port from the VMs (which is reported when the tool starts up) is not working, then the data won't be there.

When you create an AI ----- login to one of them and verify that you can reach the port (usually something like 27017) on the IP address of the host running the tool.

If it doesn't work, then that's the cause of your problem. (I prefer to avoid security groups altogether by using OpenVPN support, but that's an entirely different conversation).

/*
 * Michael R. Hines
 * Platform Engineer, DigitalOcean.
 */

Yuna Joung

unread,
Jul 29, 2016, 7:12:03 PM7/29/16
to cbtool-users, jo...@college.harvard.edu
How do you verify that you can reach the port on the IP address of the host? Could I have a specific command?


Michael R. Hines

unread,
Jul 29, 2016, 7:31:11 PM7/29/16
to Yuna Joung, cbtool-users
Either telnet or nmap will work. Those are both pretty reliable ways to
probe a port's reachability.

You can also do your own combination of netperf/iperf/ pick your
favorite network benchmark and tell it to use the port you want to be
tested.

- Michael

/*
* Michael R. Hines
* Platform Engineer, DigitalOcean.
*/

Reply all
Reply to author
Forward
0 new messages