CMU v5.1.2 and BL460 Gen8 monitoring

[Ronny Horgen]

Hi, we have SSH to and from nodes ok, but some fields show up as "No time to perform" instead of real data.

cpuload, disk_write/read, net_out/in and ib0_out/in just show "No time to perform" - all other fields are okey.

 
We run SLES 11 SP2 on the nodes.

See picture. Anyone with a good suggestion?

[Dave-SBS · Accepted solution]

Typically "no time to perform" means collectl does not yet have enough data to display yet.  I think it could also mean that the collectl daemon on that node may not be set up to collect disk, ethernet, and interconnect data.  can you post the daemon line from that node in /etc/collectl.conf.  If you  look at the manpage for collectl, under the section for subsystems you see the various parts of a system you can monitor; lowercase is summary information, and uppercase is detailed information.

[Ronny Horgen]

We don't use collectl , and on an old shelf with bl460g6 and SLES 11 SP1 it works fine with just the standard monitoring in CMU.

[Dave-SBS]

Sorry. My bad.  You don't say if you did this or not, so I will ask. Did you:

1. reinstall monitoriing on the gen8 nodes?

2. stop and restart monitoring?

[Ronny Horgen]

1. yes, I have tried this several time
2. yes, I have done this.
No change - still showing "No time to perform" for the fields indicated.

[Ronny Horgen]

Hi again, you maybe are right. It looks like collectl are set up on the head-node and first part of the cluster. I will check more on this tomorrow, and report back.

[Ronny Horgen]

Thanks Dave for getting me in the right direction. I installed and configured collectl on the new nodes, and now everything is fine.

[Dave-SBS]

Well, good; I'm glad you got that fixed. Collectl is a great tool to use both in and outside of CMU.